Title of Invention	A METHOD AND SYSTEM FOR REDIRECTING TO A FIRST SITE
Abstract	A first site comprises a first ISP, a first server and a first router interposed between the first ISP and the first server. A second site comprises a second ISP, a second server and a second router interposed between the second ISP and the second server. While the second server is operating, both the first and second routers broadcast announcements of an IP address of the second server. The announcements of the IP address broadcast by the first router indicate the first router as a less preferred route to the IP address than the announcements of the IP address by the second router indicating the second router as a route to the IP address. Consequently, packets addressed to the IP address are routed to the second server via the second ISP and the second router. Subsequently, the second site fails, and concurrently, the second router ceases to broadcast announcements of the IP address of the second site and the first router continues to broadcast announcements of the IP address of the second site. Consequently, subsequent packets addressed to the IP address are routed to the first server via the first ISP and the first router, bypassing the second ISP. A semiautomatic technique is also disclosed.

Title of Invention

A METHOD AND SYSTEM FOR REDIRECTING TO A FIRST SITE

Abstract

A first site comprises a first ISP, a first server and a first router interposed between the first ISP and the first server. A second site comprises a second ISP, a second server and a second router interposed between the second ISP and the second server. While the second server is operating, both the first and second routers broadcast announcements of an IP address of the second server. The announcements of the IP address broadcast by the first router indicate the first router as a less preferred route to the IP address than the announcements of the IP address by the second router indicating the second router as a route to the IP address. Consequently, packets addressed to the IP address are routed to the second server via the second ISP and the second router. Subsequently, the second site fails, and concurrently, the second router ceases to broadcast announcements of the IP address of the second site and the first router continues to broadcast announcements of the IP address of the second site. Consequently, subsequent packets addressed to the IP address are routed to the first server via the first ISP and the first router, bypassing the second ISP. A semiautomatic technique is also disclosed.

Full Text	SYSTEM, METHOD AND PROGRAM FOR RE-ROUTING INTERNET PACKETS r:ELD OF THE INVENTION The present invention relates generally to networks and computer systems, and more specifically to re-routing of network packets to an alternate (or recovery) site. BACKGROUND OF THE INVENTION Many types of computer networks are known today, such as Local Area Networks ("LANs"), Wide Area Networks ("WANs"), intranets, and the Internet. For example, clients make requests via the Internet to servers which reside on LANs which are connected to the Internet. It is common for one or more Internet Service Providers ("ISPs") to be logically interposed between the Internet and the LAN of a server, and for one or more edge routers to be physically and/or logically interposed between this LAN of the server and the ISP(s). Having more than one connection to the Internet is called "multihoming". The use of two or more ISPs allows for load balancing and increased resiliency. The edge routers for the server periodically broadcast Border Gateway Protocol ("BGP") announcements of the server network's Autonomous System Number '("ASN") and associated routes leading to the server. BGP is defined in PvFC 1771, and is an exterior gateway routing protocol used to share inform.ation between routers, or groups of routers, to determine efficient paths. The adjacent ISPs and routers receive these broadcasts. During normal operation, each ISP receives packets that include a source Internet Protocol ("IP") address and a destination IP address, and then forwards or "routes" the packets to the destination IP address via the intervening router (s) to the server. (A packet is a piece of a message transmitted over a packet-sv7itching network. One of the key features of a packet is that it contains the destination address in addition to the data. In IP networks, packets are often called datagrams.) The destination IP address may lead to a single server which can handle the request, a load balancer or proxy server for a pool of servers to handle the request or a gateway for the network on which the server or server pool resides. The servers at the destination IP address are sometimes called a "site" or "production site". If the site furnishes web pages to the requester as an interface to the requested application, the destination server can also be called a "website". Often, there is a backup production site ("backup site") on a backup network in case the original production site ("original site") fails, "here may be one or more ISPs for the backup site as well. When the original site fails, the edge routers of the original site stop.- their periodic BGP (Border Gateway Protocol) announcements of the original site Autonomous System Number ("ASN") and associated routes. The ISPs of the original site notice the cessation of the ASN and its associated routes for the original site, and in response, update their routing table to remove any routes associated with this ASN. The ISPs of the original site propagate these changes throughout the Internet via BGP. At this time the original site's routes and IP addresses are unknown to the Internet. To re-route packets to the backup site, it was known to use Domain Name Systems ("DNS") updates. (DNS is a system used to translate host and domain names to IP addresses.) These DNS updates change the IP address-hostname and IP address-domiain name translations. Because this solution relies on hostnames and domain names, it does not support applications or implementations that require the original IP addresses to be maintained at the backup site. Additionally, DNS updates may take up to seventy two hours to propagate worldwide (depending on individual DNS Tim.e To Live timout settings) . Extra work is required to change IP addresses at the backups site and configure DNS properly (e.g. zone file modification) . This extra work may translate to longer periods of outages. Another known solution is to install the same ISP at both sites. However, this may be costly depending on the ISP and the site(s) . If the need arises to recover at an unplanned location it is likely that the same ISP may not be installed or available at this location in a timeiy fashion. Accordingly, an object of the present invention is to expedite and facilitate re-routing of network packets to a backup site when an original site fails. SUMMARY OF THE INVENTION The present invention resides in a system, method and program for redirecting to a first site, packets addressed to a second site after failure of the second site. The first site comprises a first ISP, a first server and a first router interposed between the first ISP and the first server. The second site comprises a second ISP, a second server and a second router interposed between the second ISP and the second server. While the second server is operating, both the first and second routers broadcast announcements of an IP address of the second server. The announcements of the IP address broadcast by the first router indicate the first router as a less preferred route to the IP address than the announcements of the IP address by the second router indicating th,e second router as a route to the IP address. Consequently, packets addressed to the IP address are routed to the second server via the second ISP and the second router. Subsequently, the second site fails, and concurrently, the second router ceases to broadcast announcements of the IP address of the second site and the first router continues to broadcast announcements of the IP address of the second site. Consequently, subsequent packets addressed to the IP address are routed to the first server via the first ISP and the first router, bypassing the second ISP. According to features of the present invention, the first router is an edge router for a first network containing the first server, and the second router is an edge router for a second network containing the second server. The announcements of the IP address broadcast by the first router indicate a less preferred route to the first router than to the second router to reach the IP address. For example, the announcements of the IP address broadcast by the first router may include AS_PATn prepends to indicate a longer path to the first router than to the second router to reach the IP address. The present invention also resides in another system., method and program for redirecting to a first site, packets addressed to a second site after failure of the second site. The first site comprises a first ISP, a first server and a first router interposed between the first ISP and the first server. The second site com.prises a second ISP, a second server and a second router interposed between the second ISP and the second server. Before failure of the second site, BGP filters in the first ISP and the first router are configured to accept an IP address of the second site. While the second server is operating, the second router broadcasts announcements of the IP address of the second server, such that packets addressed to the IP address are routed to the second server via the second ISP and the second router, bypassing the first ISP. Subsequently, the second site fails, and the second router ceases to broadcast announcements of the IP address of the second site and the first router begins to broadcast announcements of the IP address of the second site, such that subsequent packets addressed to the IP address are routed to the first server via the first ISP and the first router, bypassing the second ISP. Viewed from a first aspect, the present invention provides a method for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP arid said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said method comprising the steps of: while said second server is operating, both said first and said second routers broadcasting announcem.ents of an. IP address of said second server, 'said announcements of said IP address broadcast by said first router indicating said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicating said second router as a route to said IP address, such that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and subsequently, said second site failing, and concurrently, said second router ceasing to broadcast announcements of said IP address of said second site and said first router continuing to broadcast announcements of said IP address of said second site, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP. Preferaibly, the present invention provides for a m.ethod wherein said first router is an edge router for a first network containing said first server, and said second router is an edge router for a second network containing said second server. Preferably, the present invention provides a method wherein said announcements of said IP address broadcast by said first router indicate a longer path to said first router than to said second router to reach said IP address. Preferably, the present invention provides a method wherein said announcements of said IP address broadcast by said first router include A3_PATn prepends to indicate a longer path to said first router than to said second router to reach said IP address. Preferably, the present invention provides a method further comprising the steps of: said first router also broadcasting announcements of a first ASN associated with said first site; and said second router also broadcasting announcements of a second ASN associated with said second site. Prefera.biy, the present invention provides a method further comprising the steps of: for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said method comprising the steps of: before fsilure of said second site, configuring BGP filters in said first ISP and said first router to accept an IP address of said second site before failure of said second site; while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router; subsequently, said second site failing, and said second router ceasing to broadcast announcements of said IP address and said first router broadcasting announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP. Preferably, the present invention provides a method wherein before failure of said second site, further comprising the step of configuring BGP filters in said second ISP and said second router to accept said IP address, such that while said second server is operating, packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP. j Preferably, the present invention provides a method wherein the steps of said second router ceasing to broadcast announcements of said IP address of said second site and said first router broadcasting announcem:ents of said IP address of said second site are manualiy initiated. Preferably, the present invention provides a method wherein said first router is a first edge router, and said second router is a second edge router. Preferably, the present invention provides a method further comprising the steps of: while said second server is operating, sard second router also broadcasting announcements of a second ASN associ.ated v/ith said second site; and after failure of said second site, said first router broadcasting announcements of a first ASN associated with said first site. Viewed from a second aspect the present invention provides a system for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said system comprising: means, within said first router, for broadcasting announcements of an IP address of said second server while said second server is operating; means, within said second router, for broadcasting announcements of said IP address "while said second server is operating, said announcements of said IP address broadcast by said first router indicating said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicating said second router as a route to said IP address, sucli that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and means, responsive to said second site failing, for said second router ceasing to broadcast announcements of said IP address while said first router continuing to broadcast announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP. Preferably, the present invention provides a system, wherein said first router is an edge router for a first network containing said first server, and said second router is an edge router for a second network containing said second server. Preferably, the present invention provides a system, wherein said announcemients of said IP address broadcast by said first router indicate a longer path to said first router than to said second router to reach said IP address. Preferably, the present invention provides a system -wherein said announcements of said IP address broadcast by said first router include AS_PATn prepends to indicate a longer path to said first router that to said second router to reach said IP address. Preferably, the present invention provides a system wherein: said first router also includes means for broadcasting announcements of a first ASM associated with said first site; and said second router also includes means for broadcasting announcements of a second ASN associated with said second site. Viewed from a third aspect the present invention provides a computer program product for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first IGP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said program product comprising: a computer readable media; first program, instructions, for execution within said first router, to broadcast announcements of an IP address of said second server while said second server is operating; second program instructions, for execution within said second router, to broadcast announcements of said IP address wliile said second server is operating, wherein said announcements of said IP address broadcast by said first router indicate said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicate said second router as a route to said IP address, such that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and third program instructions, responsive to said second site failing, to cause said second router to cease broadcasting announcements of said IP address while said first router continues to broadcast announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP; and wherein said first, second and third program instructions are stored on said media in functional form. Preferably, the present invention provides a computer prograrr. product as set forth in claim 10 wherein said announcements of Said IP address broadcast by said first router indicate a less preferred route to said first router than to said second router to reach said IP address. Viewed from another aspect, the present invention provides a methed for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said method comprising the steps of: before failure of said second site, configuring BGP filters in said first ISP and said first router to accept an IP address of said second site; while said second server is operaiting, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP; subsequently, said second site failing, and said second router ceasing to broadcast announcem.ents of said IP address and said first router broadcasting announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP. Preferably, the present invention provides a method wherein before failure of said second site, further comprising the step of configuring BGP filters in said second ISP and said second router to accept said IP address, such that while said second server is operating, packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP. Preferably, the present invention provides a method wherein the steps of said second router ceasing to broadcast announcements of said IP address of said second site and said first router broadcasting announcements of said IP address of said second site are manually initiated. Preferably, the present invention provides a method wherein said first router is a first edge router, and said second router is a second edge router. Preferably, the present invention provides for a method further comprising the steps of: while said second server is operating, said second router also broadcasting announcements of a second ASN associated with said second site; and after fairure of said second site, said first router broadcasting announcements of a first ASN associated with said first site. Viewed from, another aspect the present invention provides a system for redirectirig to a first server, packets addressed to a second server after failure of said second server, said systen-. com.prising: a first ISP, said first ISP having a BGP filter configured, before failure of said second server, to accept an IP address of said second server; a first router interposed between said first ISP and said first server, said first router having a BGP filter configured, before failure of said second server, to accept said IP address, said first router configured not to broadcast announcements of said IP address while said second server is operating; a second ISP, said second ISP having a BGP filter configured, during operation of said second server, to accept said IP address; a second router interposed between said second ISP and said second server, said second router having a BGP filter configured, during operation of said second server, to accept said IP address, said second router configured to broadcast announcements of said IP address while said second server is operating, such that while said second server is operating, packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP; subsequently, said second site failing, and said second router reconfigured to cease broadcasting announcements of said IP address and said first router reconfigured to broadcast announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP. Preferably, the present invention provides a system, wherein said first router is a first edge router, and said second router is a second edge router. Preferably, the present invention provides a system wherein: said second router is configured to broadcast announcements of a second ASN associated with said second server while said second server is operating; and said first router is configured to broadcast announcements of a first ASM asssociated with said first server, after but not before failure of said second server. Viewed from another aspect, the present invention provides a computer program, loadable into the internal memory of a digital computer, comprising software code portions for performing, when said product is run on a computer, to carry out the steps of the invention as described above. BRIEF DESCRIPTION OF THE FIGURES Figure 1 is a block diagram of a distributed computer system, including 'an original production site, a backup production site and edge routers associated with the original site and backup site, which embodies the present invention. Figures 2(A), 2(B) and 2(C) form, a flow chart of BGP protocol program; functions within the edge routers associated with the backup production site, and other related process steps. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will now be described in detail with reference to the figures. Figure 1 illustrates a distributed computer system, generally designated 10. Distributed computer system 10 comprises a client 12 connected to the Internet 14 via a WAN or other network 1,3. The Internet comprises a multitude of network devices including firewalls "F", routers "R", gateways "G", and switching nodes "N", as known in the prior art. System: 10 also includes two (or more) Internet Service Providers ("ISPs") 16 and 18 interposed between the Internet 14 and an original production site ("original site") 20. System. 10 also includes two (or more) Internet Service Providers ("ISPs") 46 and 48 interposed between the Internet 14 and a backup production site ("backup site") ,30. As known in the prior art, each ISP includes a known BGP filter (or equivalent) which controls which ranges of IP addresses announced from, the original site 20 or from, the backup site 30 are recognized/processed by the ISP. Thus, for each packet that reaches an ISP from its servers and from the Internet, the ISP will only process and pass the packet if its source IP address is within the range recognized and processed by the BGP filter (or equivalent) of the ISP. By way of exam.pie, the original site 20 comprises a firewall 21, a LAN network 26, one or m.ore servers 22 and 24, load balancer 25, storage 23, and other computer-related devices on network 26. The firewall 21 and load balancer 25 are optional. The use of two or more ISPs provides "multihoming",, allowing for load balaricing and increased resiliency, although if desired, only one ISP can be used. Likewise, the use of two or more servers at each site is optional, and only one server is needed at each site. Edge router 2 6 is physically and/or logically interposed between ISP 16 and firewall 21. Edge router 28 is physically and/or logically interposed between ISP 18 and firewall 21. Each ISP 16 and 18 receives packets with the destination IP address leading to a requested server 22 or 24, and then forwards or "routes" each packet to the destination IP address via the respective edge router. In the illustrated example, the destination IP address leads to load balancer 25 via firewall 21. (Alternately, the destination address is the server 22 or 24.) The load balancer 25 then for'wards the request to one of the servers 22 or 24 based on a known load balancing algorithm. (The load balancer, firewall, and associated algorithm. are not required for the present invention.) Also in the illustrated example, the request results in server 22 or 24 furnishing a V7eb page, file or other data to the requesting client 12. ISPs 16 and 18 and edge routers 26 and 28 include respective, known BGP protocol program functions (including respective address range filters 126 and 128 or equivalent) which control which ranges of source IP addresses from the original site 20 are recognized/processed by the respective ISPs and edge routers. Thus, for each packet that reaches the ISPs 16 and 18 and edge routers 26 and 28, the respective BGP protocol program function will only recognize and process the packet if its source IP address is within the range accepted by the BGP filter (or equivalent). The ov7ner of the original site 20 has also provided a backup site 30 on a backup network 36 in case the original site 20 fails. There are two or more ISPs 46 and 48 for the backup site 30. In the illustrated example, the backup site 30 includes a firewall 31, LAN network 36, a load balancer 35, two or m.ore servers 32 and 34, storage 33 and other computer-related devices on the network 36. Backup site 30 also includes edge router 36 interposed between ISP 46 and firewall 36, and edge router interposed between ISP 4 8 and firewall 36. ISPs 4 6 and 48 and edge routers 36 and 38 include respective, known BGP protocol program, functions 146, 148, 136 and 138 (including respective address range filters) which control which ranges of source IP addresses are recognized/processed by the respective ISPs and edge routers. Thus, for each packet that reaches the ISPs 46 and 48 and edge routers 36 and 38 from its servers and from; the Internet, the respective BGP protocol program, function will only recognize, process and pass the packet if its source IP address rs within the range accepted by the BGP filter (or equivalent). Based on a current inter-ISP BGP filtering standard, the IP addresses to be rerouted have a size /24 (256 contiguous IP address block) or larger in order to be received by ISP 46 and 48, announced to the rest of the Internet 14 by ISP 46 and 48, and accepted by the rest of the Internet 14 including ISP 16 and 18. However, as the inter-ISP BGP filtering standard changes, this size limitation may change, and the present invention is applicable to future network address size limitations resulting from future inter-ISP BGP filtering standards, and new version{s) of IP (e.g. IPv6) or DGP or other events. The following is a semiautomatic process for re-routing packets, originally intended for the original site 20, to the backup site 30, in accordance with one embodiment of the present invention. When original site 20 fails, the BGP protocol program, functions 116 and 118 within the edge routers 26 and 28 of the ISPs 16 and 18 learn of the failure by no longer receiving the announcements for the routes from, edge routers 2 6 and 28 (BGP protocol functions 126 and 128). Also, when the original site 20 fails, a support person learns of the problem, from, users (i.e. complaints via a help desk or problemi tickets) and notifies an admnistrator. In response, the BGP protocol functions 116, 118, 126, and 128 stop their periodic announcements of the Autonomous System. Number ("ASN") path and associated routes that represent site 20. In the illustrated example, this is the ASN path and associated routes of the original site 20 which includes servers 22 and 24, firewall 26, load balancer 25 and storage 23. The ISPs 15 and 18 notice the cessation of the announcements of the ASN path and associated routes for the original site 20, and in response, update their routing tables to rem.ove any associated route entries. ALSO, ISPs 16 and 18 notify their BGP neighbors that these routes associated with this ASN no longer exist. This information propagates through the Internet 14, world-wide including to ISPs 46 and 48 and edge routers 36 and 38, typically in less than five m.inutes. This is called Internet routing table convergence. In response to previous requests from the backup site 30 (and as preparation for a possible outage), network engineers at ISP 46 and 48 updated their BGP filters (or equivalent) to accept the announcement of IP address of the original site 20 instead from the backup site 30. In response to the failure of the original site 20, the admdnistrator at the backup site 30 configures the BGP protocol functions 136 and 138 in the edge routers 36 and 38 in the backup site 30 to broadcast announceents of the IP addresses and routes or the original site 20, using backup site BO'S ASN, to the backup ISPs 46 and 48 that are directly connected and BGP peering to the backup site 30. (BGP peering comprises exchange of BGP protocol information between two routers (peers) configured as BGP neighbors.) This tells the ISPs 46 and 48 that the backup site has a path to the IP addresses of the original site 20. ISPs 4 6 and 4 8 forward on these announcements to the rest of the Internet via BGP protocol process 146 and 148. The Internet routing table converges quickly, typically less than five minutes. Consequently, any ISPs or routers (in the Internet as well as ISPs 16, 18, 4 6, 48, and edge routers 2 6, 28, 3 6 and 38) that receive subsequent packets on the Internet addressed to the IP address(es) of the original site 20 will route these packets instead to the backup site 30. The new routing information indicates that packets addressed to the original site 2 0 should be routed directly to the ISPs 4 6 and 4 8 of the backup site 30 (without first being routed to the ISPs 16 and 18 of the original site 20) . Once the routing tables have converged, all Internet traffic intended for the original site 20 will be routed to the backup site 30 via ISPs 46 and 46 and edge routers 36 and 38 (without passing through ISPs 16 or 18). The following steps implement the foregoing re-routing process: Preliminary Steps, i.e. before failure of original site 20: a) All BGP filters and other security features (i.e. access control lists, route-m.aps, community strings, other) at edge routers 36 and 38 (and any other edge routers connected to the backup site) of ISP 46 and 48 of the backup site 30, are opened/expanded to admit address ranges associated with the original site 20. The edge routers 36 and 38 do not announce the IP address of the original site 20 to the ISPs 4 6 and 48 until there is an outage at the original site 20.. So, during nonnal operation, the other ISPs and routers do not route packets with the IP address of the original site 20 to ISPs 46 or 48. instead, the other ISPs and routers route packets with the IP address of the original site 20 to ISPs 16 or 18 en route to the original site 20. If there are any problems with implementing the readiness of the ISPs 46 and 48, edge routers 36 and 38 and backup site 30 (according to step a) above) to backup original site 20, an administrator will check BGP protocol functions 146, 148, 136 and 138 which will show the problem,. Later Step during Backup Operation, i.e. after failure of oriqinal site 20: b) After failure of the original site 20, an administrator notifies BGP protocol functions 12 6 and 12 8 of the original site 2 0 to stop announcing the IP addres:s of the original site (assuming the edge routers 2 6 and 2 8 are still operational; if they are not operational there may be no need to modify P.GP protocol functions 126 and 128). Thus, for each address range of the original site 20, the BGP protocol functions 136 and 138 in the edge routers of the backup site 30 (automatically or semi-automatically) configure a predeterm.ined EGP "announcement" that the destination IP address for the original site 20 now corresponds to the ASN and routes for the backup site 30. c) Using this announcement, the backup site 30's edge routers 36 a:nd 38 announce to the ISPs 46, and 48 at backup site 30, changes to source addresses that should be routed through the BGP filters of the ISPs 4 6 and 48 (and any other ISPs that are directly connected and BGP peering to the backup site 30). Using this announcement, the backup site 30's edge routers also announce to the ISPs 46, and 48 (and any other ISPs that are directly connected and BGP peering to backup site 30) that the backup site 30 will receive the packets addressed to the original site 20. (ISPs typically filter routing announcements that they will accept or to which they will listen at the /24 (256 contiguous IP addresses) range level. Also, this minimizes the size of Internet routing tables.) In the illustrated example, the address range being recovered is a full /24 (256 contiguous IP addresses) or larger range to be accepted or listened to by BGP filters (or equivalent) within the BGP protocol functions 116, 118, 126, 128, 146, 148, 136 and 138 as well as any other Internet routers worldwide accepting /24 or larger announcements. As an example, a /23 network has a larger range or block of IP addresses than a /24 network. f) After production site 20 is recovered/restored, the BGP announcements of step b) and c) from the edge routers 36 and 38 of the backup site 30 are terminated, and the admnistrator configures original site 20 EGP protocol process 126 and 128 to once again announce the original IP addresses of original site 20 and their associated routes via its ASN to the ISPs 16 and 18 of the original site 20. ISPs 16 and 18 then announce to the rest of the world that they should receive and process packets addressed to the servers 22 and 24, i.e. so that ISPs 16 and 18 will receive and pass through packets addressed to the original site 2 0 and ISPs 4 6 and 4 8 will not. The following is an automatic process for re-routing packets in accordance with another embodiment of the present invention. when original site 20 fails, the BGP protocol functions 12 6 and 128 within edge routers 26 and 28 stop their periodic announcements of the Autonomous System. Num.ber ("ASM") path and routes of original site 20. The ISPs 26 and 18 notice the cessation of the ASN path and corresponding routes for the original site 20, and in response, update their routing table to remove the routes associated with this ASN path and notify their BGP neighbors who in turn notify their BGP neighbors until the, Internet routing table reconverges that these specific routes to this ASN no longer exist. BGP protocol functions 136 and 138 for the backup) site 30 have also been broadcasting BGP announcements for these same routes but these broadcasts are indicated as less preferred than the broadcasts from the original site routes. When the broadcasts of the more preferred routes from; original site 20 stop, the broadcasts of the routes from, backup site 30 will take precedence. Once the Internet routing table reconverges with these less preferred routes, any ISPs or routers that receive subsequent packets on the Internet addressed to the IP address of the original site 20 will route these packets to the backup site 30. The nevv routing information indicates that packets addressed to the original site 20 should be routed directly to the ISPs 46 and 48 of the backup site 30, without first, being routed to the ISPs 15 and 18 of the original site 20. Once the routing tables have converged, all Internet traffic intended for the original site 20 will be routed directly to the backup site 30 (without passing through ISPs 16 or 18) . in this embodiment of the present invention, upon failure of original site 20 and cessation of the announcements from. BGP protocol functions 126 and 128, the announcements by BGP protocol functions 136 and 138 automatically take preference. This is im.plemented by making the routes announced by the backup site 30 ASN routers 3 6 and 38 less preferred than the routes announced by the original site ASN routers 2 6 and 28. Therefore, when the original site 20 is operating normally it will receive all traffic destined for the original site 20 IP addresses, allowever, during a failure of the original site 20 and its ability to announce its routes to the Internet 14, the routes broadcast from backup site 30 become preferred and propagate through the worldwide Internet 14. By way of example, the BGP protocol functions 136 and 138 can use AS_PATn prepends to make their announcements less preferred than the announcem.ents from, the original site 20 BGP protocol functions 126 and 128. The AS path length is the fifth item in the BGPv4 route selection decision process. By appending the backup site 30 ASN multiple times onto the AS_PAll this lengthens the AS_PATII to reach the backup site 30. This way the original site 20 has no AS_PATII prepends (in an effort to have the sh.ortest, and most preferred AS_PATII possible) while the backup, site 30 has multiple AS_PATll prepends lengthening its A3 path thus making it less desirable, or _ess preferred, by BGP. Therefore, under normal operating modes, EGP will select the original site 20 -with the shortest AS path length until such time as the original site 20 fails and the backup site 30 with the longer AS path length becomes the more desirable, or preferred, route. Six AS_PATn prepends are currently recommended to make this effective. Figures 2(A), 2(B) and 2(C) illustrate the foregoing automatic and semiautomatic processes in more detail. Preferably, the following steps 100-140 and 170-174 are performed as setup prior to failure of original site 20. Tills will (identify any problems in the setup and) prepare backup site 30 to perform its backup role in the event of subsequent failure of original site 20. In step 100, an admdnistrator identifies the IP address of the original site and the ASN of the original site and the ASN of the backup site 30, and configures each of the BGP protocol functions 136 and 138. Next, the admdnistrator determdnes if the IP address space of the original site 20 is longer than the BGP filters in ISPs 46 and 48, in current day practice (/24, or 256 contiguous IP addresses or larger, under today's practice) and can therefore be rerouted through the Internet (decision 102). Each of the ISPs 46 and 48 includes a BGP filter, or equivalent, which determines which announcements and packets can pass through the ISPs to the Internet 14. If the IP address range of the original site 20 is not a large enough block of IP addresses (currently /24 or larger) (decision 102, no branch), then the IP addresses and packets addressed to the original site 20 cannot be redirected to the backup site 30 via this methodology, and no further action can be taken by each of the BGP protocol functions 136 and 138, except to notify an admdnistrator of the problem (step 104). Referring again to decision 102, yes branch, if the IP address space to be redirected meets the BGP filter (or equivalent) requirements, then the administrator and ISP 16, 18, 46, and 48 determdne if the backup site 30 is authorized to receive packets addressed to the original site 20 (decision 106). This determination is based on administrator and ISP (16, 18, 46, and 48) authorizations and enforced via ISP and admdnistrator filters in the BGP"protocol functions 116, 118, 146, 148, 126, 128, 136, and 138. Typically, an admdnistrator sets these configuration files in edge routers 26, 28, 36 and 38. If backup site 30 is not authorized to backup original site 20 (decision 106;, no branch), then the packets addressed to the original site 20 cannot be redirected to the backup site 30, and no further action is taken by each of the EGP protocol functions 13G and _38, except to notify an administrator of the problem (step 108). Alternately, the administrator may change the authorization to authorize backup site 30 to backup original site 20. If backup site 30 is authorized to backup original site 20, then the administrator determines if the backup site 30 has its own Autonomous System Number ("ASN") (decision 110) . If not (decision 110, no branch), for one ISP (e.g. ISP 4 5) connection only the ISP 4 6 assigns a private ASN to the backup site 30 (step 114) and the administrator configures BGP protocol functions 136 and 138 with the ASN. For a multihomed environment, the administrator may obtain an ASN from, the Americar Reegistry for Internet Numbers (www.arin.net) and continue from, decision 110, yes branch. Next, each of the BGP protocol functions 136 and 138 notifies ISPs 16 and 18 and ISPs 46 and 48 of their intent to redirect packets addressed to original site 20 instead to backup site 30, and the ASN for backup site 30, as follows (step 120) . (Each edge router and ISP broadcasts its IP address and the ASN of the network containing the IP address. The ASNs uniquely identify respective networks (containing one or m.ore IP address) , and are used to facilitate routing of the packet from the Internet to the destination IP address.) The edge routers 36 and 38 broadcast/announce the IP address of the original site 20, but their broadcast/announcement will be in a secondary or "less preferred" manner, such that the recipient Internet routers of the broadcast/announcem.ent will still route packets addressed to the IP address of the original site to ISPs 16 and 18, and ISPs 16 and 18 will still route these packets to the original site 20 as long as edge routers 26 and 28 continue to broadcast/announce their routes to the IP address of the original site 20. This is because the broadcast/announcemient of the IP address of the original site 20 by edge routers 26 and 28 will be in a primary or "more preferred" manner as long as the original site 20 is active and edge routers 26 and 28 continue to announce the IP address of the original site. Next, the administrator of the BGP protocol functions 136 and 138 receives a notification from ISPs 46 and 48 whether ISPs 46 and 4 8 will update their BGP filters, access lists and other security features (i.e. access control lists, other) to redirect packets addressed to original site 20 instead to backup site 30 (decision 124). ISPs 46 and 4 8 m.ake this decision based on technical feasibility and internal customer support decisions. (It has already been determined in decisions 102 and 106 that the address space can be technically redirected and that backup site 30 is authorized to backup original site 20.) If ISPs 46 and 48 will not redirect packets addressed to original site 20 instead to backup site 30 (decision 124, no branch), then BGP protocol functions 136 and 138 cannot redirect these packets and notify an administrator accordingly (step _26). however, If ISPs 46 and 48 will redirect packets addressed to original site 20 instead to backup site 30 (decision 124, yes branch), then the administrator requests that ISPs 46 and 48 update their BGP filters, access lists and other security features to redirect packets addressed to original site 20 instead to backup:) site 30 (step 130) . ISPs 46 and 48 comply. Next, each of the BGP protocol functions 136 and 138 checlcs its configuration file to determine whether BGP protocol functions 136 and 138 are configured to automiatically generate the notification to all ISPs (not just ISPs 16, 18, 46 and 48)and all routers that are directly connected and connected by BGP peering to edge routers 36 and 38 to route packets addressed to original site 20 instead to backup site 30 (decision 140). Semiautomatic Announcements If the BGP protocol functions 136 amd 138 are not configured to automatically generate the notifications (decision 140, no branch) , then each of the BGP protocol functions 136 and 138 sets a flag (clescription or other) to indicate an intent for manual generation of the notification and waits for failure of the original site 20 (step 142). Upon subsequent failure of the original site 20 (step 146), an administrator will notice the failure and (a) set each of the BGP protocol functions 126 and 128 to stop announcing the IP address and ASN of the original site 20 to the ISPs 16 and 18 (step 150) and (b) set each of the BGP protocol functions 136 and 138 to generate a notification to ISPs 46 and 48 and any other ISPs and routers that are directly connected and BGP peering to routers 3 6 and 38, to redirect all packets addressed to original site 20 instead to backup site 30 (step 158). In response, each of the BGP protocol functions 138 and 138 generates and broadcasts the announcement (step 158) . Upon subsequent restoration of the original site 20 (step 159), the administrator (a) sets the BGP protocol functions 136 and 138 to stop announcing the IP address of the original site 2 0 (step 160) and (b) sets the BGP protocol functions 126 and 128 to start announcing the IP address of the original site 20 to the ISPs 16 and 18 and any other ISPs and routers that are directly connected or connected by BGP peering to edge routers 26 and 28 (step 162) . In response, new packets addressed to the original site 20 are routed to ISPs 16 and 18 of the original site and then via edge routers 26 or 28 to original site 20 (step 164) . Automatic Announcements Refer again to decision 14 0, yes branch where the BGP protocol iunctions 126, 128, 136 and 138 are configured for automatic operation. In such a case, the administrator of the BGP protocol functions 136 and 138 determines if the backup site 30 has its own ASN (decision 170). (The admiinistrator can obtain an ASN from, the American Registry for Internet Numbers www.arin.net.) If not (decision 170, no branch), then the BGP pjrotocol functions 136 and 138 do not perform the automatic notification of the redirect, and instead notify an administrator to perform the foregoing semiautomatic process of steps 150-160 and proceeds to step 172 to set the flag indicating semiautomatic notification (step 172) . allwever, if the backup site 30 has or can obtain its own ASN (decision 170, yes branch) , then each of the BGP protocol functiorns 126, 128, 136 and 138 automatically prepares the announcements of the redirect based on a setting in a configuration file within edge routers 26, 28, 36, and 38 (step- 171) , and sets a flag to indicate ready for automatic announcement of the redirect (step; 173) . By way of example, the BGP protocol functions 136 and 138 of the backup site can use AS__PATI1 prepends to make their announcem.ents of the IP address of the original site 20 less preferred than the announcements by BGP protocol functions 126 and 128 of the original site 20 (step 174). The AS path length is the fifth item in the BGPv4 decision process. The AS_PATII prepend step comprises appending the backup site 30 ASN multiple times onto the AS_PATI1. This lengthens the AS_PATII to reach the backup site 30. This way the original site 20 has no AS_PATII prepends (in an effort to have the shortest, and m.ost preferred AS_PATII possible) while the backup site 30 has m.ultiple AS_PATII prepends lengthening its AS path thus making it less desirable, or less preferred, by BGP. Therefore, under normal operating modes, BGP will select the original site 2 0 with the shortest AS path length until such time as the original site 20 fails and the backup site 30 becomes the more desirable, or preferred, route. Six AS_PATI1 prepends are currently recommended to make this effective. Thurs, each of the BGP protocol functions 136 and 138 makes their (less preferred) announcements to ISPs 4 6 and 48 of the IP address of the original site 20, that packets addressed to the original site 20 should be routed instead directly to the backup site 30 (step 174). Packets with the IP address of the original site 20 continue r.o route to the original site 20 until there is an outage of the original site 20. If and when the original site 20 fails, its edge routers 26 and 28 cease to announce to IS)Ps 16 and 18 and to all other routers that are directly connected and connected by BGP peering to edge routers 2 6 and 28, the IP address of the original site 20 (step 178) . In response, the announcements by PGP protocol functions 136 and 138 to ISPs 46 and 48 and all other routers that are directly connected and connected by BGP peering to edge routers 36 and 38 (that packets addressed to the original site 20 should instead be routed directly to the backup") site 30) become the primary/controlling announcement. Consequently, subsequent packets addressed to the IP address of original site 20 will route directly to backup site 30 (step 179) . When the original site 20 is subsequently restored, the edge routers 26 and 28 once again announce to ISPs 16 and 18, and all other routers that are directly connected and connected by BGP peering to edge routers 26 and 28 the IP address of the original site 20 (step 180). In response, BGP protocol functions 14 6 and 148 will cease to accept and cease to insert into the routing table as preferred routes, the announcements from BGP protocol 136 and 138 the IP address of the original site 2 0 (step 182), and subsequent packets will route to ISPs 16 and 18 and then to the original site 2 0 via edge routers 2 6 and 2 8 (without passing through ISPs 46 and 48 or edge routers 36 and 38 (step 184) . BGP protocol functions 136 and 138 in functional form (for example, executable form.) can be loaded into edge routers 36 and 38 from, computer readable media 236 and 238 such as m.agnetic disk or tape, optical disk, DVD, semiconductor memory, etc. or downloaded from the Internet or wireless network via TCP/IP or other network adapter cards in edge routers 36 and 38. BGP protocol functions 126 and 128 in functional form, (for example, executable form.) can be loaded into edge routers 2 6 and 2 8 from computer readable media 236 and 238 such as magnetic disk or tape, optical disk, DVD, semiiconductor memory, etc. or downloaded from the Internet or wireless network via TCP/IP or other network adapter cards in edge routers 26 and 28. Eased on the foregoing, system., method and program, product for redirecting packets to a backup site have been disclosed. However, numerous modifications and substitutions can be made without deviating from the scope of the present invention. For example; other means of making routes less or more preferred may be used by edge routers 3 6 and 38 includirig, but not limited to, community Strang exchange and action upori the community string; BGP filter masks (/24) may change length in the future without change to the rdea presented above. Therefore, the present invention has been disclosed by way of illustration and not limitation, and reference should be made to the following claims to determ.ine the scope of the present invention. CLAIMS 1. A method for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said method comprising the steps of: while said second server is operating, both said first and said second routers broadcasting announcements of an IP address of said second server, said announcements of said IP address broadcast by said first router indicating said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicating said second router as a route to said IP address, such that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and subsequently, said second site failing, and concurrently, said second router ceasing to broadcast announcements of said IP address of said second site and said first router continuing to broadcast announcements of said IP address of said second site, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP. 2. A method as claimed in claim 1 wherein said first router is an edge router for a first network containing said first server, and sa.id second router is an edge router for a second network containing said second server. 3. A method as claimed in claim 1 wherein said announcements of said IP address broadcast by said first router indicate a longer path to said first router than to said second router to reach said IP address. 4. A method as claimed in claim 1 wherein said announcements of said IP address broadcast by said first router include As_PATH prepends to indicate a longer path to said first router than to said second router to reach said IP address. 5. A method as claimed in claim 1 further comprising the steps of: said first router also broadcasting announcements of a first ASM associated with said first site; and said second router also broadcasting announcements of a second ASM associated with said second site. 6. A method as claimed in claim 1 further comprising the steps of: redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server: and a second router interposed between said second ISP and said second server, said method comprising the steps of: before failure of said second site, configuring BGP filters in said first ISP and said first router to accept an IP address of said second site before failure of said second site; while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router; subsequently, said second site failing, and said second router ceasing to broadcast announcements of said IP address and said first router broadcasting announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP. 7. A method as claimed in claim 6 wherein before failure of said second site, further comprising the step of configuring BGP filters in said second ISP and said second router to accept said IP address, such that while said second server is operating, packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP. 8. A method as claimed in claim 7 wherein the steps of sa.id second router ceasing to broadcast announcements of said IP address of said second site and said first router broadcasting announcements of said jp address of said second site are manually initiated. 9. A method as claimed in claim 6 wherein said first router is a tirsr edge router, and said second router is a second edge router. 10. A method as claimed in claim 9 further comprising the steps of: while said second server is operating, said second router also broadcasting announcements of a second ASN, associated with said second site; and after failure of said second site, said first router broadcasting announcements of a first ASN associated with said first site. 11. A system for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said system comprising configured to perform the method as claimed in any of the preceding claims 1 to 10. 12. A product loadable into the internal memory on computer, when executed on the computer is configured to perform the method as claimed in claims 1 to 10.

Full Text

SYSTEM, METHOD AND PROGRAM FOR RE-ROUTING INTERNET PACKETS
r:ELD OF THE INVENTION
The present invention relates generally to networks and computer systems, and more specifically to re-routing of network packets to an alternate (or recovery) site.
BACKGROUND OF THE INVENTION
Many types of computer networks are known today, such as Local Area Networks ("LANs"), Wide Area Networks ("WANs"), intranets, and the Internet. For example, clients make requests via the Internet to servers which reside on LANs which are connected to the Internet. It is common for one or more Internet Service Providers ("ISPs") to be logically interposed between the Internet and the LAN of a server, and for one or more edge routers to be physically and/or logically interposed between this LAN of the server and the ISP(s). Having more than one connection to the Internet is called "multihoming". The use of two or more ISPs allows for load balancing and increased resiliency. The edge routers for the server periodically broadcast Border Gateway Protocol ("BGP") announcements of the server network's Autonomous System Number '("ASN") and associated routes leading to the server. BGP is defined in PvFC 1771, and is an exterior gateway routing protocol used to share inform.ation between routers, or groups of routers, to determine efficient paths. The adjacent ISPs and routers receive these broadcasts. During normal operation, each ISP receives packets that include a source Internet Protocol ("IP") address and a destination IP address, and then forwards or "routes" the packets to the destination IP address via the intervening router (s) to the server. (A packet is a piece of a message transmitted over a packet-sv7itching network. One of the key features of a packet is that it contains the destination address in addition to the data. In IP networks, packets are often called datagrams.) The destination IP address may lead to a single server which can handle the request, a load balancer or proxy server for a pool of servers to handle the request or a gateway for the network on which the server or server pool resides. The servers at the destination IP address are sometimes called a "site" or "production site". If the site furnishes web pages to the requester as an interface to the requested application, the destination server can also be called a "website".

Often, there is a backup production site ("backup site") on a backup network in case the original production site ("original site") fails, "here may be one or more ISPs for the backup site as well. When the original site fails, the edge routers of the original site stop.- their periodic BGP (Border Gateway Protocol) announcements of the original site Autonomous System Number ("ASN") and associated routes. The ISPs of the original site notice the cessation of the ASN and its associated routes for the original site, and in response, update their routing table to remove any routes associated with this ASN. The ISPs of the original site propagate these changes throughout the Internet via BGP. At this time the original site's routes and IP addresses are unknown to the Internet. To re-route packets to the backup site, it was known to use Domain Name Systems ("DNS") updates. (DNS is a system used to translate host and domain names to IP addresses.) These DNS updates change the IP address-hostname and IP address-domiain name translations. Because this solution relies on hostnames and domain names, it does not support applications or implementations that require the original IP addresses to be maintained at the backup site. Additionally, DNS updates may take up to seventy two hours to propagate worldwide (depending on individual DNS Tim.e To Live timout settings) . Extra work is required to change IP addresses at the backups site and configure DNS properly (e.g. zone file modification) . This extra work may translate to longer periods of outages. Another known solution is to install the same ISP at both sites. However, this may be costly depending on the ISP and the site(s) . If the need arises to recover at an unplanned location it is likely that the same ISP may not be installed or available at this location in a timeiy fashion.
Accordingly, an object of the present invention is to expedite and facilitate re-routing of network packets to a backup site when an original site fails.
SUMMARY OF THE INVENTION
The present invention resides in a system, method and program for redirecting to a first site, packets addressed to a second site after failure of the second site. The first site comprises a first ISP, a first server and a first router interposed between the first ISP and the first server. The second site comprises a second ISP, a second server and a second router interposed between the second ISP and the second server. While the second server is operating, both the first and second routers

broadcast announcements of an IP address of the second server. The announcements of the IP address broadcast by the first router indicate the first router as a less preferred route to the IP address than the announcements of the IP address by the second router indicating th,e second router as a route to the IP address. Consequently, packets addressed to the IP address are routed to the second server via the second ISP and the second router. Subsequently, the second site fails, and concurrently, the second router ceases to broadcast announcements of the IP address of the second site and the first router continues to broadcast announcements of the IP address of the second site. Consequently, subsequent packets addressed to the IP address are routed to the first server via the first ISP and the first router, bypassing the second ISP.
According to features of the present invention, the first router is an edge router for a first network containing the first server, and the second router is an edge router for a second network containing the second server. The announcements of the IP address broadcast by the first router indicate a less preferred route to the first router than to the second router to reach the IP address. For example, the announcements of the IP address broadcast by the first router may include AS_PATn prepends to indicate a longer path to the first router than to the second router to reach the IP address.
The present invention also resides in another system., method and program for redirecting to a first site, packets addressed to a second site after failure of the second site. The first site comprises a first ISP, a first server and a first router interposed between the first ISP and the first server. The second site com.prises a second ISP, a second server and a second router interposed between the second ISP and the second server. Before failure of the second site, BGP filters in the first ISP and the first router are configured to accept an IP address of the second site. While the second server is operating, the second router broadcasts announcements of the IP address of the second server, such that packets addressed to the IP address are routed to the second server via the second ISP and the second router, bypassing the first ISP. Subsequently, the second site fails, and the second router ceases to broadcast announcements of the IP address of the second site and the first router begins to broadcast announcements of the IP address of the second site, such that subsequent packets addressed to the IP address are routed to the first server via the first ISP and the first router, bypassing the second ISP.

Viewed from a first aspect, the present invention provides a method for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP arid said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said method comprising the steps of: while said second server is operating, both said first and said second routers broadcasting announcem.ents of an. IP address of said second server, 'said announcements of said IP address broadcast by said first router indicating said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicating said second router as a route to said IP address, such that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and subsequently, said second site failing, and concurrently, said second router ceasing to broadcast announcements of said IP address of said second site and said first router continuing to broadcast announcements of said IP address of said second site, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP.
Preferaibly, the present invention provides for a m.ethod wherein said first router is an edge router for a first network containing said first server, and said second router is an edge router for a second network containing said second server.
Preferably, the present invention provides a method wherein said
announcements of said IP address broadcast by said first router indicate a longer path to said first router than to said second router to reach said IP address.
Preferably, the present invention provides a method wherein said announcements of said IP address broadcast by said first router include A3_PATn prepends to indicate a longer path to said first router than to said second router to reach said IP address.
Preferably, the present invention provides a method further
comprising the steps of: said first router also broadcasting announcements

of a first ASN associated with said first site; and said second router also broadcasting announcements of a second ASN associated with said second site.
Prefera.biy, the present invention provides a method further comprising the steps of: for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said method comprising the steps of: before fsilure of said second site, configuring BGP filters in said first ISP and said first router to accept an IP address of said second site before failure of said second site; while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router; subsequently, said second site failing, and said second router ceasing to broadcast announcements of said IP address and said first router broadcasting announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP.
Preferably, the present invention provides a method wherein before failure of said second site, further comprising the step of configuring BGP filters in said second ISP and said second router to accept said IP address, such that while said second server is operating, packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP. j
Preferably, the present invention provides a method wherein the steps of said second router ceasing to broadcast announcements of said IP address of said second site and said first router broadcasting announcem:ents of said IP address of said second site are manualiy initiated.

Preferably, the present invention provides a method wherein said first router is a first edge router, and said second router is a second edge router.
Preferably, the present invention provides a method further comprising the steps of: while said second server is operating, sard second router also broadcasting announcements of a second ASN associ.ated v/ith said second site; and after failure of said second site, said first router broadcasting announcements of a first ASN associated with said first site.
Viewed from a second aspect the present invention provides a system for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said system comprising: means, within said first router, for broadcasting announcements of an IP address of said second server while said second server is operating; means, within said second router, for broadcasting announcements of said IP address "while said second server is operating, said announcements of said IP address broadcast by said first router indicating said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicating said second router as a route to said IP address, sucli that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and means, responsive to said second site failing, for said second router ceasing to broadcast announcements of said IP address while said first router continuing to broadcast announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP.
Preferably, the present invention provides a system, wherein said
first router is an edge router for a first network containing said first server, and said second router is an edge router for a second network containing said second server.
Preferably, the present invention provides a system, wherein said
announcemients of said IP address broadcast by said first router indicate a

longer path to said first router than to said second router to reach said IP address.
Preferably, the present invention provides a system -wherein said
announcements of said IP address broadcast by said first router include
AS_PATn prepends to indicate a longer path to said first router that to
said second router to reach said IP address.
Preferably, the present invention provides a system wherein: said first router also includes means for broadcasting announcements of a first ASM associated with said first site; and said second router also includes means for broadcasting announcements of a second ASN associated with said second site.
Viewed from a third aspect the present invention provides a computer program product for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first IGP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said program product comprising: a computer readable media; first program, instructions, for execution within said first router, to broadcast announcements of an IP address of said second server while said second server is operating; second program instructions, for execution within said second router, to broadcast announcements of said IP address wliile said second server is operating, wherein said announcements of said IP address broadcast by said first router indicate said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicate said second router as a route to said IP address, such that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and third program instructions, responsive to said second site failing, to cause said second router to cease broadcasting announcements of said IP address while said first router continues to broadcast announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP; and wherein said first, second and third program instructions are stored on said media in functional form.

Preferably, the present invention provides a computer prograrr. product as set forth in claim 10 wherein said announcements of Said IP address broadcast by said first router indicate a less preferred route to said first router than to said second router to reach said IP address.
Viewed from another aspect, the present invention provides a methed for redirecting to a first site, packets addressed to a second site after failure of said second site, said first site comprising a first ISP, a first server and a first router interposed between said first ISP and said first server, said second site comprising a second ISP, a second server and a second router interposed between said second ISP and said second server, said method comprising the steps of: before failure of said second site, configuring BGP filters in said first ISP and said first router to accept an IP address of said second site; while said second server is operaiting, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP; subsequently, said second site failing, and said second router ceasing to broadcast announcem.ents of said IP address and said first router broadcasting announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP.
Preferably, the present invention provides a method wherein before failure of said second site, further comprising the step of configuring BGP filters in said second ISP and said second router to accept said IP address, such that while said second server is operating, packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP.
Preferably, the present invention provides a method wherein the steps of said second router ceasing to broadcast announcements of said IP address of said second site and said first router broadcasting announcements of said IP address of said second site are manually initiated.
Preferably, the present invention provides a method wherein said first router is a first edge router, and said second router is a second edge router.

Preferably, the present invention provides for a method further comprising the steps of: while said second server is operating, said second router also broadcasting announcements of a second ASN associated with said second site; and after fairure of said second site, said first router broadcasting announcements of a first ASN associated with said first site.
Viewed from, another aspect the present invention provides a system for redirectirig to a first server, packets addressed to a second server after failure of said second server, said systen-. com.prising: a first ISP, said first ISP having a BGP filter configured, before failure of said second server, to accept an IP address of said second server; a first router interposed between said first ISP and said first server, said first router having a BGP filter configured, before failure of said second server, to accept said IP address, said first router configured not to broadcast announcements of said IP address while said second server is operating; a second ISP, said second ISP having a BGP filter configured, during operation of said second server, to accept said IP address; a second router interposed between said second ISP and said second server, said second router having a BGP filter configured, during operation of said second server, to accept said IP address, said second router configured to broadcast announcements of said IP address while said second server is operating, such that while said second server is operating, packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP; subsequently, said second site failing, and said second router reconfigured to cease broadcasting announcements of said IP address and said first router reconfigured to broadcast announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP.
Preferably, the present invention provides a system, wherein said first router is a first edge router, and said second router is a second edge router.
Preferably, the present invention provides a system wherein: said second router is configured to broadcast announcements of a second ASN associated with said second server while said second server is operating; and said first router is configured to broadcast announcements of a first

ASM asssociated with said first server, after but not before failure of said second server.
Viewed from another aspect, the present invention provides a computer program, loadable into the internal memory of a digital computer, comprising software code portions for performing, when said product is run on a computer, to carry out the steps of the invention as described above.
BRIEF DESCRIPTION OF THE FIGURES Figure 1 is a block diagram of a distributed computer system, including 'an original production site, a backup production site and edge routers associated with the original site and backup site, which embodies the present invention.
Figures 2(A), 2(B) and 2(C) form, a flow chart of BGP protocol program; functions within the edge routers associated with the backup production site, and other related process steps.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
The present invention will now be described in detail with reference to the figures. Figure 1 illustrates a distributed computer system, generally designated 10. Distributed computer system 10 comprises a client 12 connected to the Internet 14 via a WAN or other network 1,3. The Internet comprises a multitude of network devices including firewalls "F", routers "R", gateways "G", and switching nodes "N", as known in the prior art. System: 10 also includes two (or more) Internet Service Providers ("ISPs") 16 and 18 interposed between the Internet 14 and an original production site ("original site") 20. System. 10 also includes two (or more) Internet Service Providers ("ISPs") 46 and 48 interposed between the Internet 14 and a backup production site ("backup site") ,30. As known in the prior art, each ISP includes a known BGP filter (or equivalent) which controls which ranges of IP addresses announced from, the original site 20 or from, the backup site 30 are recognized/processed by the ISP. Thus, for each packet that reaches an ISP from its servers and from the Internet, the ISP will only process and pass the packet if its source IP address is within the range recognized and processed by the BGP filter (or equivalent) of the ISP. By way of exam.pie, the original site 20 comprises a firewall 21, a LAN network 26, one or m.ore servers 22 and 24, load balancer 25, storage 23, and other computer-related devices on network 26. The firewall 21 and load balancer 25 are optional. The use of two or more

ISPs provides "multihoming",, allowing for load balaricing and increased resiliency, although if desired, only one ISP can be used. Likewise, the use of two or more servers at each site is optional, and only one server is needed at each site. Edge router 2 6 is physically and/or logically interposed between ISP 16 and firewall 21. Edge router 28 is physically and/or logically interposed between ISP 18 and firewall 21. Each ISP 16 and 18 receives packets with the destination IP address leading to a requested server 22 or 24, and then forwards or "routes" each packet to the destination IP address via the respective edge router. In the illustrated example, the destination IP address leads to load balancer 25 via firewall 21. (Alternately, the destination address is the server 22 or 24.) The load balancer 25 then for'wards the request to one of the servers 22 or 24 based on a known load balancing algorithm. (The load balancer, firewall, and associated algorithm. are not required for the present invention.) Also in the illustrated example, the request results in server 22 or 24 furnishing a V7eb page, file or other data to the requesting client 12. ISPs 16 and 18 and edge routers 26 and 28 include respective, known BGP protocol program functions (including respective address range filters 126 and 128 or equivalent) which control which ranges of source IP addresses from the original site 20 are recognized/processed by the respective ISPs and edge routers. Thus, for each packet that reaches the ISPs 16 and 18 and edge routers 26 and 28, the respective BGP protocol program function will only recognize and process the packet if its source IP address is within the range accepted by the BGP filter (or equivalent).
The ov7ner of the original site 20 has also provided a backup site 30 on a backup network 36 in case the original site 20 fails. There are two or more ISPs 46 and 48 for the backup site 30. In the illustrated example, the backup site 30 includes a firewall 31, LAN network 36, a load balancer 35, two or m.ore servers 32 and 34, storage 33 and other computer-related devices on the network 36. Backup site 30 also includes edge router 36 interposed between ISP 46 and firewall 36, and edge router interposed between ISP 4 8 and firewall 36. ISPs 4 6 and 48 and edge
routers 36 and 38 include respective, known BGP protocol program, functions 146, 148, 136 and 138 (including respective address range filters) which control which ranges of source IP addresses are recognized/processed by the respective ISPs and edge routers. Thus, for each packet that reaches the ISPs 46 and 48 and edge routers 36 and 38 from its servers and from;
the Internet, the respective BGP protocol program, function will only

recognize, process and pass the packet if its source IP address rs within the range accepted by the BGP filter (or equivalent).
Based on a current inter-ISP BGP filtering standard, the IP
addresses to be rerouted have a size /24 (256 contiguous IP address block)
or larger in order to be received by ISP 46 and 48, announced to the rest of the Internet 14 by ISP 46 and 48, and accepted by the rest of the Internet 14 including ISP 16 and 18. However, as the inter-ISP BGP filtering standard changes, this size limitation may change, and the present invention is applicable to future network address size limitations resulting from future inter-ISP BGP filtering standards, and new version{s) of IP (e.g. IPv6) or DGP or other events.
The following is a semiautomatic process for re-routing packets, originally intended for the original site 20, to the backup site 30, in accordance with one embodiment of the present invention. When original site 20 fails, the BGP protocol program, functions 116 and 118 within the edge routers 26 and 28 of the ISPs 16 and 18 learn of the failure by no longer receiving the announcements for the routes from, edge routers 2 6 and 28 (BGP protocol functions 126 and 128). Also, when the original site 20 fails, a support person learns of the problem, from, users (i.e. complaints via a help desk or problemi tickets) and notifies an admnistrator. In response, the BGP protocol functions 116, 118, 126, and 128 stop their periodic announcements of the Autonomous System. Number ("ASN") path and associated routes that represent site 20. In the illustrated example, this is the ASN path and associated routes of the original site 20 which includes servers 22 and 24, firewall 26, load balancer 25 and storage 23. The ISPs 15 and 18 notice the cessation of the announcements of the ASN path and associated routes for the original site 20, and in response, update their routing tables to rem.ove any associated route entries. ALSO, ISPs 16 and 18 notify their BGP neighbors that these routes associated with this ASN no longer exist. This information propagates through the Internet 14, world-wide including to ISPs 46 and 48 and edge routers 36 and 38, typically in less than five m.inutes. This is called Internet routing table convergence.
In response to previous requests from the backup site 30 (and as preparation for a possible outage), network engineers at ISP 46 and 48 updated their BGP filters (or equivalent) to accept the announcement of IP address of the original site 20 instead from the backup site 30. In response to the failure of the original site 20, the admdnistrator at the backup site 30 configures the BGP protocol functions 136 and 138 in the

edge routers 36 and 38 in the backup site 30 to broadcast announceents of the IP addresses and routes or the original site 20, using backup site BO'S ASN, to the backup ISPs 46 and 48 that are directly connected and BGP peering to the backup site 30. (BGP peering comprises exchange of BGP protocol information between two routers (peers) configured as BGP neighbors.) This tells the ISPs 46 and 48 that the backup site has a path to the IP addresses of the original site 20. ISPs 4 6 and 4 8 forward on these announcements to the rest of the Internet via BGP protocol process 146 and 148. The Internet routing table converges quickly, typically less than five minutes. Consequently, any ISPs or routers (in the Internet as well as ISPs 16, 18, 4 6, 48, and edge routers 2 6, 28, 3 6 and 38) that receive subsequent packets on the Internet addressed to the IP address(es) of the original site 20 will route these packets instead to the backup site 30. The new routing information indicates that packets addressed to the original site 2 0 should be routed directly to the ISPs 4 6 and 4 8 of the backup site 30 (without first being routed to the ISPs 16 and 18 of the original site 20) . Once the routing tables have converged, all Internet traffic intended for the original site 20 will be routed to the backup site 30 via ISPs 46 and 46 and edge routers 36 and 38 (without passing through ISPs 16 or 18).
The following steps implement the foregoing re-routing process:
Preliminary Steps, i.e. before failure of original site 20:
a) All BGP filters and other security features (i.e. access control lists, route-m.aps, community strings, other) at edge routers 36 and 38 (and any other edge routers connected to the backup site) of ISP 46 and 48 of the backup site 30, are opened/expanded to admit address ranges associated with the original site 20. The edge routers 36 and 38 do not announce the IP address of the original site 20 to the ISPs 4 6 and 48 until there is an outage at the original site 20.. So, during nonnal operation, the other ISPs and routers do not route packets with the IP address of the original site 20 to ISPs 46 or 48. instead, the other ISPs
and routers route packets with the IP address of the original site 20 to ISPs 16 or 18 en route to the original site 20. If there are any problems with implementing the readiness of the ISPs 46 and 48, edge routers 36 and 38 and backup site 30 (according to step a) above) to backup original site 20, an administrator will check BGP protocol functions 146, 148, 136 and
138 which will show the problem,.

Later Step during Backup Operation, i.e. after failure of oriqinal site 20:
b) After failure of the original site 20, an administrator notifies BGP
protocol functions 12 6 and 12 8 of the original site 2 0 to stop announcing
the IP addres:s of the original site (assuming the edge routers 2 6 and 2 8
are still operational; if they are not operational there may be no need
to modify P.GP protocol functions 126 and 128). Thus, for each address
range of the original site 20, the BGP protocol functions 136 and 138 in
the edge routers of the backup site 30 (automatically or
semi-automatically) configure a predeterm.ined EGP "announcement" that the destination IP address for the original site 20 now corresponds to the ASN and routes for the backup site 30.
c) Using this announcement, the backup site 30's edge routers 36 a:nd 38
announce to the ISPs 46, and 48 at backup site 30, changes to source
addresses that should be routed through the BGP filters of the ISPs 4 6 and
48 (and any other ISPs that are directly connected and BGP peering to the
backup site 30). Using this announcement, the backup site 30's edge
routers also announce to the ISPs 46, and 48 (and any other ISPs that are
directly connected and BGP peering to backup site 30) that the backup site
30 will receive the packets addressed to the original site 20. (ISPs
typically filter routing announcements that they will accept or to which
they will listen at the /24 (256 contiguous IP addresses) range level.
Also, this minimizes the size of Internet routing tables.) In the
illustrated example, the address range being recovered is a full /24 (256
contiguous IP addresses) or larger range to be accepted or listened to by
BGP filters (or equivalent) within the BGP protocol functions 116, 118,
126, 128, 146, 148, 136 and 138 as well as any other Internet routers
worldwide accepting /24 or larger announcements. As an example, a /23
network has a larger range or block of IP addresses than a /24 network.
f) After production site 20 is recovered/restored, the BGP announcements of step b) and c) from the edge routers 36 and 38 of the backup site 30 are terminated, and the admnistrator configures original
site 20 EGP protocol process 126 and 128 to once again announce the
original IP addresses of original site 20 and their associated routes via its ASN to the ISPs 16 and 18 of the original site 20. ISPs 16 and 18 then announce to the rest of the world that they should receive and process packets addressed to the servers 22 and 24, i.e. so that ISPs 16
and 18 will receive and pass through packets addressed to the original site 2 0 and ISPs 4 6 and 4 8 will not.

The following is an automatic process for re-routing packets in accordance with another embodiment of the present invention. when original site 20 fails, the BGP protocol functions 12 6 and 128 within edge routers 26 and 28 stop their periodic announcements of the Autonomous System. Num.ber ("ASM") path and routes of original site 20. The ISPs 26 and 18 notice the cessation of the ASN path and corresponding routes for the original site 20, and in response, update their routing table to remove the routes associated with this ASN path and notify their BGP neighbors who in turn notify their BGP neighbors until the, Internet routing table reconverges that these specific routes to this ASN no longer exist. BGP protocol functions 136 and 138 for the backup) site 30 have also been broadcasting BGP announcements for these same routes but these broadcasts are indicated as less preferred than the broadcasts from the original site routes. When the broadcasts of the more preferred routes from; original site 20 stop, the broadcasts of the routes from, backup site 30 will take precedence. Once the Internet routing table reconverges with these less preferred routes, any ISPs or routers that receive subsequent packets on the Internet addressed to the IP address of the original site 20 will route these packets to the backup site 30. The nevv routing information indicates that packets addressed to the original site 20 should be routed directly to the ISPs 46 and 48 of the backup site 30, without first, being routed to the ISPs 15 and 18 of the original site 20. Once the routing tables have converged, all Internet traffic intended for the original site 20 will be routed directly to the backup site 30 (without passing through ISPs 16 or 18) .
in this embodiment of the present invention, upon failure of original site 20 and cessation of the announcements from. BGP protocol functions 126 and 128, the announcements by BGP protocol functions 136 and 138 automatically take preference. This is im.plemented by making the routes announced by the backup site 30 ASN routers 3 6 and 38 less preferred than the routes announced by the original site ASN routers 2 6 and 28. Therefore, when the original site 20 is operating normally it will receive all traffic destined for the original site 20 IP addresses, allowever, during a failure of the original site 20 and its ability to announce its routes to the Internet 14, the routes broadcast from backup site 30 become preferred and propagate through the worldwide Internet 14. By way of example, the BGP protocol functions 136 and 138 can use AS_PATn prepends to make their announcements less preferred than the announcem.ents from, the original site 20 BGP protocol functions 126 and 128. The AS path

length is the fifth item in the BGPv4 route selection decision process. By appending the backup site 30 ASN multiple times onto the AS_PAll this lengthens the AS_PATII to reach the backup site 30. This way the original site 20 has no AS_PATII prepends (in an effort to have the sh.ortest, and most preferred AS_PATII possible) while the backup, site 30 has multiple AS_PATll prepends lengthening its A3 path thus making it less desirable, or _ess preferred, by BGP. Therefore, under normal operating modes, EGP will select the original site 20 -with the shortest AS path length until such time as the original site 20 fails and the backup site 30 with the longer AS path length becomes the more desirable, or preferred, route. Six AS_PATn prepends are currently recommended to make this effective.
Figures 2(A), 2(B) and 2(C) illustrate the foregoing automatic and semiautomatic processes in more detail. Preferably, the following steps 100-140 and 170-174 are performed as setup prior to failure of original site 20. Tills will (identify any problems in the setup and) prepare backup site 30 to perform its backup role in the event of subsequent failure of original site 20. In step 100, an admdnistrator identifies the IP address of the original site and the ASN of the original site and the ASN of the backup site 30, and configures each of the BGP protocol functions 136 and 138. Next, the admdnistrator determdnes if the IP address space of the original site 20 is longer than the BGP filters in ISPs 46 and 48, in current day practice (/24, or 256 contiguous IP addresses or larger, under today's practice) and can therefore be rerouted through the Internet (decision 102). Each of the ISPs 46 and 48 includes a BGP filter, or equivalent, which determines which announcements and packets can pass through the ISPs to the Internet 14. If the IP address range of the original site 20 is not a large enough block of IP addresses (currently /24 or larger) (decision 102, no branch), then the IP addresses and packets addressed to the original site 20 cannot be redirected to the backup site 30 via this methodology, and no further action can be taken by each of the BGP protocol functions 136 and 138, except to notify an admdnistrator of the problem (step 104). Referring again to decision 102, yes branch, if the IP address space to be redirected meets the BGP filter (or equivalent) requirements, then the administrator and ISP 16, 18, 46, and 48 determdne if the backup site 30 is authorized to receive packets addressed to the original site 20 (decision 106). This determination is based on administrator and ISP (16, 18, 46, and 48) authorizations and enforced via ISP and admdnistrator filters in the BGP"protocol functions 116, 118, 146, 148, 126, 128, 136, and 138. Typically, an admdnistrator sets these configuration files in edge routers 26, 28, 36 and 38. If

backup site 30 is not authorized to backup original site 20 (decision 106;, no branch), then the packets addressed to the original site 20 cannot be redirected to the backup site 30, and no further action is taken by each of the EGP protocol functions 13G and _38, except to notify an administrator of the problem (step 108). Alternately, the administrator may change the authorization to authorize backup site 30 to backup original site 20. If backup site 30 is authorized to backup original site 20, then the administrator determines if the backup site 30 has its own Autonomous System Number ("ASN") (decision 110) . If not (decision 110, no branch), for one ISP (e.g. ISP 4 5) connection only the ISP 4 6 assigns a private ASN to the backup site 30 (step 114) and the administrator configures BGP protocol functions 136 and 138 with the ASN. For a multihomed environment, the administrator may obtain an ASN from, the Americar Reegistry for Internet Numbers (www.arin.net) and continue from, decision 110, yes branch. Next, each of the BGP protocol functions 136 and 138 notifies ISPs 16 and 18 and ISPs 46 and 48 of their intent to redirect packets addressed to original site 20 instead to backup site 30, and the ASN for backup site 30, as follows (step 120) . (Each edge router and ISP broadcasts its IP address and the ASN of the network containing the IP address. The ASNs uniquely identify respective networks (containing one or m.ore IP address) , and are used to facilitate routing of the packet from the Internet to the destination IP address.) The edge routers 36 and 38 broadcast/announce the IP address of the original site 20, but their broadcast/announcement will be in a secondary or "less preferred" manner, such that the recipient Internet routers of the broadcast/announcem.ent will still route packets addressed to the IP address of the original site to ISPs 16 and 18, and ISPs 16 and 18 will still route these packets to the original site 20 as long as edge routers 26 and 28 continue to broadcast/announce their routes to the IP address of the original site 20. This is because the broadcast/announcemient of the IP address of the original site 20 by edge routers 26 and 28 will be in a primary or "more preferred" manner as long as the original site 20 is active and edge routers 26 and 28 continue to announce the IP address of the original site. Next, the administrator of the BGP protocol functions 136 and 138 receives a notification from ISPs 46 and 48 whether ISPs 46 and 4 8 will update their BGP filters, access lists and other security features (i.e. access control lists, other) to redirect packets addressed to original site 20 instead to backup site 30 (decision 124). ISPs 46 and 4 8 m.ake this decision based on technical feasibility and internal customer support decisions. (It has already been determined in decisions 102 and 106 that the address space can be technically redirected and that backup

site 30 is authorized to backup original site 20.) If ISPs 46 and 48 will not redirect packets addressed to original site 20 instead to backup site 30 (decision 124, no branch), then BGP protocol functions 136 and 138 cannot redirect these packets and notify an administrator accordingly (step _26). however, If ISPs 46 and 48 will redirect packets addressed to original site 20 instead to backup site 30 (decision 124, yes branch), then the administrator requests that ISPs 46 and 48 update their BGP filters, access lists and other security features to redirect packets addressed to original site 20 instead to backup:) site 30 (step 130) . ISPs 46 and 48 comply. Next, each of the BGP protocol functions 136 and 138 checlcs its configuration file to determine whether BGP protocol functions 136 and 138 are configured to automiatically generate the notification to all ISPs (not just ISPs 16, 18, 46 and 48)and all routers that are directly connected and connected by BGP peering to edge routers 36 and 38 to route packets addressed to original site 20 instead to backup site 30 (decision 140).
Semiautomatic Announcements
If the BGP protocol functions 136 amd 138 are not configured to automatically generate the notifications (decision 140, no branch) , then each of the BGP protocol functions 136 and 138 sets a flag (clescription or other) to indicate an intent for manual generation of the notification and waits for failure of the original site 20 (step 142). Upon subsequent failure of the original site 20 (step 146), an administrator will notice the failure and (a) set each of the BGP protocol functions 126 and 128 to stop announcing the IP address and ASN of the original site 20 to the ISPs 16 and 18 (step 150) and (b) set each of the BGP protocol functions 136 and 138 to generate a notification to ISPs 46 and 48 and any other ISPs and routers that are directly connected and BGP peering to routers 3 6 and 38, to redirect all packets addressed to original site 20 instead to backup site 30 (step 158). In response, each of the BGP protocol functions 138 and 138 generates and broadcasts the announcement (step 158) .
Upon subsequent restoration of the original site 20 (step 159), the administrator (a) sets the BGP protocol functions 136 and 138 to stop announcing the IP address of the original site 2 0 (step 160) and (b) sets the BGP protocol functions 126 and 128 to start announcing the IP address of the original site 20 to the ISPs 16 and 18 and any other ISPs and routers that are directly connected or connected by BGP peering to edge

routers 26 and 28 (step 162) . In response, new packets addressed to the original site 20 are routed to ISPs 16 and 18 of the original site and then via edge routers 26 or 28 to original site 20 (step 164) .
Automatic Announcements
Refer again to decision 14 0, yes branch where the BGP protocol iunctions 126, 128, 136 and 138 are configured for automatic operation. In such a case, the administrator of the BGP protocol functions 136 and 138 determines if the backup site 30 has its own ASN (decision 170). (The admiinistrator can obtain an ASN from, the American Registry for Internet Numbers www.arin.net.) If not (decision 170, no branch), then the BGP pjrotocol functions 136 and 138 do not perform the automatic notification of the redirect, and instead notify an administrator to perform the foregoing semiautomatic process of steps 150-160 and proceeds to step 172 to set the flag indicating semiautomatic notification (step 172) . allwever, if the backup site 30 has or can obtain its own ASN (decision 170, yes branch) , then each of the BGP protocol functiorns 126, 128, 136 and 138 automatically prepares the announcements of the redirect based on a setting in a configuration file within edge routers 26, 28, 36, and 38 (step- 171) , and sets a flag to indicate ready for automatic announcement of the redirect (step; 173) . By way of example, the BGP protocol functions 136 and 138 of the backup site can use AS__PATI1 prepends to make their announcem.ents of the IP address of the original site 20 less preferred than the announcements by BGP protocol functions 126 and 128 of the original site 20 (step 174). The AS path length is the fifth item in the BGPv4 decision process. The AS_PATII prepend step comprises appending the backup site 30 ASN multiple times onto the AS_PATI1. This lengthens the AS_PATII to reach the backup site 30. This way the original site 20 has no AS_PATII prepends (in an effort to have the shortest, and m.ost preferred AS_PATII possible) while the backup site 30 has m.ultiple AS_PATII prepends lengthening its AS path thus making it less desirable, or less preferred, by BGP. Therefore, under normal operating modes, BGP will select the original site 2 0 with the shortest AS path length until such time as the original site 20 fails and the backup site 30 becomes the more desirable, or preferred, route. Six AS_PATI1 prepends are currently recommended to make this effective.
Thurs, each of the BGP protocol functions 136 and 138 makes their (less preferred) announcements to ISPs 4 6 and 48 of the IP address of the original site 20, that packets addressed to the original site 20 should

be routed instead directly to the backup site 30 (step 174). Packets with the IP address of the original site 20 continue r.o route to the original site 20 until there is an outage of the original site 20. If and when the original site 20 fails, its edge routers 26 and 28 cease to announce to IS)Ps 16 and 18 and to all other routers that are directly connected and connected by BGP peering to edge routers 2 6 and 28, the IP address of the original site 20 (step 178) . In response, the announcements by PGP protocol functions 136 and 138 to ISPs 46 and 48 and all other routers that are directly connected and connected by BGP peering to edge routers 36 and 38 (that packets addressed to the original site 20 should instead be routed directly to the backup") site 30) become the primary/controlling announcement. Consequently, subsequent packets addressed to the IP address of original site 20 will route directly to backup site 30 (step 179) . When the original site 20 is subsequently restored, the edge routers 26 and 28 once again announce to ISPs 16 and 18, and all other routers that are directly connected and connected by BGP peering to edge routers 26 and 28 the IP address of the original site 20 (step 180). In response, BGP protocol functions 14 6 and 148 will cease to accept and cease to insert into the routing table as preferred routes, the announcements from BGP protocol 136 and 138 the IP address of the original site 2 0 (step 182), and subsequent packets will route to ISPs 16 and 18 and then to the original site 2 0 via edge routers 2 6 and 2 8 (without passing through ISPs 46 and 48 or edge routers 36 and 38 (step 184) .
BGP protocol functions 136 and 138 in functional form (for example, executable form.) can be loaded into edge routers 36 and 38 from, computer readable media 236 and 238 such as m.agnetic disk or tape, optical disk, DVD, semiconductor memory, etc. or downloaded from the Internet or wireless network via TCP/IP or other network adapter cards in edge routers 36 and 38.
BGP protocol functions 126 and 128 in functional form, (for example, executable form.) can be loaded into edge routers 2 6 and 2 8 from computer readable media 236 and 238 such as magnetic disk or tape, optical disk, DVD, semiiconductor memory, etc. or downloaded from the Internet or wireless network via TCP/IP or other network adapter cards in edge routers 26 and 28.
Eased on the foregoing, system., method and program, product for redirecting packets to a backup site have been disclosed. However, numerous modifications and substitutions can be made without deviating

from the scope of the present invention. For example; other means of
making routes less or more preferred may be used by edge routers 3 6 and 38
includirig, but not limited to, community Strang exchange and action upori
the community string; BGP filter masks (/24) may change length in the
future without change to the rdea presented above. Therefore, the present
invention has been disclosed by way of illustration and not limitation,
and reference should be made to the following claims to determ.ine the
scope of the present invention.

CLAIMS
1. A method for redirecting to a first site, packets addressed to a
second site after failure of said second site, said first site comprising
a first ISP, a first server and a first router interposed between said
first ISP and said first server, said second site comprising a second ISP,
a second server and a second router interposed between said second ISP and
said second server, said method comprising the steps of:
while said second server is operating, both said first and said second routers broadcasting announcements of an IP address of said second server, said announcements of said IP address broadcast by said first router indicating said first router as a less preferred route to said IP address than said announcements of said IP address by said second router indicating said second router as a route to said IP address, such that during operation of said second server, packets addressed to said IP address are routed to said second server via said second ISP and said second router; and
subsequently, said second site failing, and concurrently, said second router ceasing to broadcast announcements of said IP address of said second site and said first router continuing to broadcast announcements of said IP address of said second site, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP.
2. A method as claimed in claim 1 wherein said first router is an edge router for a first network containing said first server, and sa.id second router is an edge router for a second network containing said second server.
3. A method as claimed in claim 1 wherein said announcements of said IP address broadcast by said first router indicate a longer path to said first router than to said second router to reach said IP address.
4. A method as claimed in claim 1 wherein said announcements of said IP address broadcast by said first router include As_PATH prepends to indicate a longer path to said first router than to said second router to reach said IP address.
5. A method as claimed in claim 1 further comprising the steps of:

said first router also broadcasting announcements of a first ASM associated with said first site; and
said second router also broadcasting announcements of a second ASM associated with said second site.
6. A method as claimed in claim 1 further comprising the steps of:
redirecting to a first site, packets addressed to a second site after
failure of said second site, said first site comprising a first ISP, a
first server and a first router interposed between said first ISP and said
first server, said second site comprising a second ISP, a second server:
and a second router interposed between said second ISP and said second
server, said method comprising the steps of:
before failure of said second site, configuring BGP filters in said first ISP and said first router to accept an IP address of said second site before failure of said second site;
while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router, bypassing said first ISP while said second server is operating, said first router not broadcasting announcements of said IP address and said second router broadcasting announcements of said IP address, such that packets addressed to said IP address are routed to said second server via said second ISP and said second router;
subsequently, said second site failing, and said second router ceasing to broadcast announcements of said IP address and said first router broadcasting announcements of said IP address, such that subsequent packets addressed to said IP address are routed to said first server via said first ISP and said first router, bypassing said second ISP.
7. A method as claimed in claim 6 wherein before failure of said second
site, further comprising the step of configuring BGP filters in said
second ISP and said second router to accept said IP address, such that
while said second server is operating, packets addressed to said IP
address are routed to said second server via said second ISP and said
second router, bypassing said first ISP.

8. A method as claimed in claim 7 wherein the steps of sa.id second router ceasing to broadcast announcements of said IP address of said second site and said first router broadcasting announcements of said jp address of said second site are manually initiated.
9. A method as claimed in claim 6 wherein said first router is a tirsr
edge router, and said second router is a second edge router.
10. A method as claimed in claim 9 further comprising the steps of:
while said second server is operating, said second router also
broadcasting announcements of a second ASN, associated with said second site; and
after failure of said second site, said first router broadcasting announcements of a first ASN associated with said first site.
11. A system for redirecting to a first site, packets addressed to a
second site after failure of said second site, said first site comprising
a first ISP, a first server and a first router interposed between said
first ISP and said first server, said second site comprising a second ISP,
a second server and a second router interposed between said second ISP and
said second server, said system comprising configured to perform the
method as claimed in any of the preceding claims 1 to 10.
12. A product loadable into the internal memory on computer, when
executed on the computer is configured to perform the method as claimed in
claims 1 to 10.

Documents:

4736-CHENP-2008 AMENDED PAGES OF SPECIFICATION 19-09-2013.pdf

4736-CHENP-2008 AMENDED CLAIMS 19-09-2013.pdf

4736-CHENP-2008 EXAMINATION REPORT REPLY RECEIVED 19-09-2013.pdf

4736-CHENP-2008 FORM-1 19-09-2013.pdf

4736-CHENP-2008 FORM-3 19-09-2013.pdf

4736-CHENP-2008 POWER OF ATTORNEY 19-09-2013.pdf

4736-CHENP-2008 PRIORITY DOCUMENT 19-09-2013.pdf

4736-CHENP-2008 CORRESPONDENCE OTHERS 29-05-2013.pdf

4736-chenp-2008 abstract.pdf

4736-chenp-2008 assignment.pdf

4736-chenp-2008 claims.pdf

4736-chenp-2008 correspondence-others.pdf

4736-chenp-2008 description (complete).pdf

4736-chenp-2008 drawings.pdf

4736-chenp-2008 form-1.pdf

4736-chenp-2008 form-18.pdf

4736-chenp-2008 form-3.pdf

4736-chenp-2008 form-5.pdf

4736-chenp-2008 others.pdf

4736-chenp-2008 pct.pdf

« Previous Patent

Next Patent »

Patent Number

263674

Indian Patent Application Number

4736/CHENP/2008

PG Journal Number

46/2014

Publication Date

14-Nov-2014

Grant Date

13-Nov-2014

Date of Filing

08-Sep-2008

Name of Patentee

INTERNATIONAL BUSINESS MACHINES CORPORATION

Applicant Address

NEW ORCHARD ROAD, ARMONK, NEW YORK 10504,

Inventors:

#	Inventor's Name	Inventor's Address
1	HOFER, SCOTT,	4460 MAROON CIRCLE, BLOOMFIELD, COLORADO 80020,

PCT International Classification Number

H04L12/28

PCT International Application Number

PCT/EP07/50602

PCT International Filing date

2007-01-22

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1	11/350,567	2006-02-09	U.S.A.