Title of Invention

DATA PROCESSING APPARATUS

Abstract

Data processing apparatus having a communication interface for connecting the data processing apparatus associated with a first network node to a network, characterized by coupling means for providing availability-increasing functions in conjunction with at least one further data processing apparatus which has coupling means, which is associated with a further network node and which is connected to the network, the network being a network of the kind in which data processing apparatuses are not registered and configurable at a central location.

Full Text

Coupling means—for a data processing apparatus The invention relates to coupling means for a data processing apparatus for coupling a first data processing apparatus (which is connected to a network) to at least one second data processing apparatus (which is likewise connected to the network, and has coupling means), with the second data processing apparatus not being identified in the network. Conventional networks are used, above all, in telecommunications. These networks are characterized in that they have "end-to-end control". That is to say, the physical connection can be established from the starting point to the end point. The status of the network elements is always clearly defined. Networks of this type have the advantage that it is possible to implement very reliable transmission paths by routing the transmission only via extremely reliable components. This is therefore achieved by virtue of the extremely high reliability of the individual network components and by virtue of a highly developed and thus expensive network management system. The disadvantage of networks of this type is, on the one hand, the high complexity and, on the other hand, a performance limitation as a result of the rigid definitions of the device parameters. This ultimately leads to a high cost/performance ratio for all [lacuna]. A further disadvantage resides in the fact that the services which can be provided in the network are defined by the network and the elements themselves and are thus available to a very limited extent only. Customer-related and flexible services cannot be provided in a network of this type. "IP networks" (IP = Internet Protocol) which also satisfy the new requirements for services which are to be provided were set up within the context of connecting previously separate networks. Said IP networks are connected to intelligent nodes, a multiplicity of servers and data stored on the latter, which can interact for the purpose of carrying out a service. Data is transmitted in packets. In this case, a volume of data to be transmitted is divided into individual packets which are transported separately. The packets are reassembled in the receiver. The data processing apparatuses are not identified in the network. A network having data processing apparatuses which are not identified is to be understood to mean those networks in which the individual data processing apparatuses are not registered at a central location where they can be configured. "Clusters", in contrast, do not fall into this network category since, in cluster networks, the individual data processing apparatuses are exactly known and configured. In this case, the exact hardware configuration and the type of systems are of fundamental importance for the operability of the data processing apparatuses which have been joined together. IP networks (which are those networks having data processing apparatuses which are not identified) result in completely new boundary conditions for providing services in networks. Services are not necessarily dependent on transport mechanisms in the network. It is possible to transmit multimedia contents. Since data does not have to be transmitted in real time, transmission may be effected using a dynamic bandwidth which may be changed even during a service. In this case, the transport of data and the service are separate from one another and may be supplied by different providers. In this case, a service may be addressed both logically and in accordance with content. However, IP networks also have disadvantages. End-to-end control within the network and the associated capability of providing high-availability connections and nodes are thus not possible since the nodes (via which the data is transported) within the network can be defined dynamically. Whereas, in classic telecommunications networks, the entire network or at least a large part of the network is owned and administered by one operator, in an IP network the network elements, services and associated contents belong to many different owners. This also means that end-to-end control is not possible. Restricting the network to only very powerful components in order to achieve high availability contradicts the openness of the network and free access to service and also the free introduction of service into the network. Cost advantages resulting from the ability to integrate any desired components would be lost. The problem with IP networks is thus that it is not possible to guarantee the high availability which is often required. Increasing the availability of an individual network node does not sufficiently increase the overall system reliability of all the components which are required or used to provide a service. The reliability of all the network components in question cannot be established since there are too many of them and, furthermore, they may be distributed throughout the world. Neither is it possible to achieve reliability by carrying out end-to-end control since the latter is not provided in an IP network and the requisite administration procedures are not available. One known solution to the problem is to adopt concepts from classic telecommunications networks and their use in IP networks. Accordingly, vulnerable intercontinental connections, for example, may be provided by a particular operator and the remaining intercontinental paths may be handled via the Internet. The selection of said particular operator by a user makes it possible for the latter to ensure that the connection via the IP network satisfies certain quality criteria. However, the flexibility of the IP network is then lost again. "Intelligent networks" are another possible solution. In these networks, the type of services which may be provided is limited. Intelligent networks are supported on central databases and form a layer which is superordinate to a telephone network. The availability of the central databases may be ensured by the use of cluster techniques at selected large service nodes. However, this results in extremely large service nodes and at best influences the operation of a single node. However, the "intelligence" of future IP-based networks will not be restricted to central nodes but will be distributed among all network components in the IP network, with the result that measures on individual nodes will be far from sufficient. A further problem resides in the fact that only the availability of the hardware is improved by equipping the hardware with functions which increase availability. However, the flexibility of a service generally depends on the software implementation, in which case it should be noted that approximately 4 0% of all faults are due to software. Purely improving the hardware situation thus only partly contributes to increasing the availability of a service. It is customary practice in e-commerce to replicate nodes one or more times in order to lessen the effects of a node failing. However, this concept, applied to a single node, is not sufficient in the case of complex services since a greater number of nodes are required to provide a service of this type. It is an object of the invention, therefore, to specify coupling means which can be used to connect a data processing apparatus to a network and which make it possible to provide services in a highly reliable manner. This object is achieved by coupling means for a data processing apparatus of the type mentioned initially which is characterized in that the coupling means, in conjunction with the second data processing apparatus, provide functions which increase availability. The inventive coupling means make it possible to equip any desired data processing apparatuses (which are not identified in a network) in such a manner that it is possible to achieve increased availability by virtue of the interaction of said data processing apparatuses. The combination of a plurality of data processing systems using inventive coupling means in order to provide functions which increase availability is referred to below as a "virtual cluster". The advantage of the inventive coupling means thus resides in the fact that it is possible to use standard hardware components which are not subject to any particular requirements. There is no need to match the hardware of the data processing apparatuses. A further advantage resides in the fact that the coupling means may be added by remote installation since they may be a pure software component. Various data processing apparatuses can therefore be provided with the inventive coupling means merely from one location in order to set up a network having data processing apparatuses which are not identified in the network. A network which has high-availability properties can thus be set up from a central location. Furthermore, in comparison with conventional clusters in which a data processing apparatus can belong to only one cluster, there is the advantage, in a "virtual cluster" having inventive coupling means, that a data processing apparatus may also belong to a plurality of_ logical clusters. Further advantageous refinements of the invention are specified in subclaims. The invention is explained in more detail below with reference to an exemplary embodiment. In the accompanying drawings: Fig. 1 shows the basic structure of an IP network when handling a service which is distributed among a plurality of nodes, Fig. 2 shows the structure of a node having a conventional cluster, Fig. 3 shows an arrangement of a network having a plurality of nodes with inventive coupling means, Figs. 4 to 7 show various phases in the handling of a service when using nodes with inventive coupling means. In the illustration in Fig. 1, a user of a user computer A, referred to below as "client", requests a service in an Ip network. The IP network is a network having nodes which are not identified, as explained in the introduction to the description. Client A's request first of all arrives at a node Kl which cannot provide the requested service on its own. Said node therefore forwards a further request to the node K2 to provide that part of the service which the node Kl cannot manage to do. The node K2 is likewise not capable of providing all the services for satisfying client A's request and therefore forwards a subtask to the node K3. A chain of 3 nodes Kl to K3 is therefore required in order to satisfy user A's request. The nodes Kl to K3 generally neither belong to a single owner nor do they comprise the same hardware and software. The reliability of the nodes Kl to K3 is thus generally very different. The node K3 could, for example, be a cluster comprising a plurality of servers, said cluster handling payment transactions at a bank in order to settle accounts for the service requested by client A. A high-availability cluster can achieve a reliability level of 99.999% in this case, corresponding to a downtime of 5 minutes per year. From client A's point of view, however, only the reliability of the entire chain comprising the three nodes Kl, K2 and K3 is important. If only the node K3 has been equipped to have a high level of availability, this does not necessarily result in a high level of availability for the entire service. It is conceivable to use a cluster solution as shown in Fig. 2 in order to make an individual node have a high level of availability. Every server in the cluster has hardware HW, an operating system OS and cluster software or hardware, respectively. The servers are connected to one another, with the cluster software/hardware controlling functions which increase availability. All of the various servers, that is to say the cluster, are then connected to a network, to an IP network in the present case. This node thus has a high level of availability by virtue of the fact that measures have been provided locally for the purpose of increasing availability. In this context, the term "locally" may also mean, for example, an inhouse network in which a plurality of servers are connected to form a cluster even though they are arranged in a spatially separate mariner. In this sense, locally is thus to be understood as meaning that the hardware and system configuration are known and have been matched to one another. Joint system and cluster administration is carried out. All data in the cluster is known and may be retrieved at a central location. Fig. 3 shows a high-availability configuration, as could be produced in accordance with the invention. There are nodes (node 1) which, as previously, have been configured to have a high level of availability by means of local measures. The availability of the node 1 can be increased further by providing additional inventive coupling means which make it possible to implement availability-increasing functions via the IP network in conjunction with other nodes. A simple node which itself does not contain a local cluster can likewise be made to have a high level of availability by adding coupling means which enable coupling to other nodes and thus make it possible to implement functions which increase availability. In the example of Fig. 3, coupling of the nodes 2 and 3 using inventive coupling means and the IP network may result in just as high an availability level as a local cluster at node 1, when considered on its own. The scope of functions of the coupling means is designed in such a manner that customary functions of a local cluster are also available for inventive virtual clusters. In the event that the coupling means are additionally used in a local cluster, functions which increase availability are available both at a local level and via the IP network. The functions which increase availability may, for example, involve providing means for memory replication to the second data processing apparatus. The memory of a data processing apparatus is thus always mapped onto the second data processing apparatus, so that, if the first data processing apparatus fails or a fault occurs therein, it is possible to fall back on the memory replica in order to restore the state before the fault occurred. A second capability of functions which increase availability involves calculations being carried out in parallel on the first data processing apparatus and on a second data processing apparatus and the result then being compared. If the results are different, it may be concluded that there is a fault and appropriate fault handling measures may be initiated. It is furthermore customary practice, as a function which increases availability, to provide timer objects in redundant form. In addition, it is possible to reciprocally monitor the addressability of the respective other data processing apparatus. This makes it possible to ascertain when one of the data processing apparatuses is no longer working correctly, and the other data processing apparatus can assume processing of the tasks to be dealt with. It is optionally possible to take action on the other data processing apparatus to such an extent as to terminate an application which is no longer working. However, it is also possible, at this point, to provide many-other functions with which the person skilled in the art is familiar and which are therefore covered by the invention. All these functions may be implemented using software. A tremendous advantage of the inventive coupling means thus resides in the fact that it is not necessary to take any significant action on the system. It is furthermore advantageous that the coupling means may be installed remotely. This is important, above all, when it is not possible to actually physically access the other data processing apparatuses, which will be the norm with an IP network such as the Internet. The manner in which the capabilities of inventive coupling means are used is explained below with reference to an instance of application. The example relates to a travel reservation service. In this case, a trip (initially not determined in any more detail) is to be booked and paid for. A user on client A sends an enquiry to a first node SX1 in an IP network (Fig. 4). This node may be, for example, the server of a service provider. The following tasks must be carried out in order to fully book a trip. - Identification of a trip by enquiring with various service providers, - selection of some trips of interest and determination of the price by accessing the corresponding service providers, - selection of a trip and execution of the booking, and - preparation of the travel documents for the customer and billing of the customer. As shown in Fig. 4, various servers of various service providers will have to collaborate in the course of carrying out these subtasks. In Fig. 5, the user on client A consults the service provider node SXl. SX1 may, for example, be the website of a tour operator. Client A is a PC which has been equipped to interchange data with the Internet. Following contact with SXl, SXl realizes that carrying out the requested travel service will place high demands on availability. Following failure of the server, said operator would possibly lose the customer and the turnover to be achieved thereby. Therefore, it must be ensured that the service proceeds in a high-availability environment. Instead of equipping the node so as to have a high level of availability, as shown in Fig. 2, the node SXl consults two neighboring nodes SU and SX2 which are known to said node SXl and are likewise able to carry out the "travel reservation" service. The manner in which the nodes which can provide the same service are recognized does not form the subject matter of the invention and is not explained any further here. One simple option is to administer services of this type using common tables which can be reciprocally updated in the event of changes. It is likewise assumed that the table lists only those nodes on which the addressed service has already been installed. However, there could also be a relationship which allows the node SXl to install the service remotely. This approach is used, for example, in peer-to-peer solutions (which are becoming widespread) when exchanging music or when jointly calculating extensive algorithms. In a step 2a, the node SXl first of all enquires of the node SU whether it is prepared to act as backup for the node SXl. The response from the node SU turns out to be negative. Possible reasons are that the node is already overloaded or has exceeded a maximum permissible access limit. Another possibility is the lack of a redundant connection between SXl and SU or an insufficient communication bandwidth. In a subsequent step 2b, the node SXl consults the node SX2 and obtains a positive response there. SXl and SX2 then enter into a virtual connection which is temporary under certain circumstances, that is to say a virtual cluster NX is formed from the two nodes SXl and SX2. This means that the "session" started with client A is replicated on the node SX2, that is to say the data interchanged between client A and SXl is replicated as contexts on the node SX2 and processes on SX2 are configured as backup processes for the same processes on the node SXl. A database which is administered via a standard interface is used for configuration. Special programming interfaces are available for this purpose. Fundamental configuration parameters for this are: - processes and their functions in the virtual cluster, - connections between the nodes (e.g. SXl and SX2) and their parameters; high-speed networks such as, for example, gigabit Ethernet or lOObaseT Ethernet are normally used between the nodes, - naming conventions of the administered objects (contexts, timers, files etc.), - monitoring parameters (e.g. waiting time until a system failure is ascertained), - audit parameters concerning the objects to be monitored and associated recovery measures, - signaling channels for signaling events and alarms, and - storage of statistical information. The following situation exists at the end of step 2. A desired trip has been identified by enquiring with various service providers. Trips of interest, for example on the basis of the prices offered, have been selected by accessing the corresponding service providers. The node SX2 has in this case been configured as the backup node for SX1 and can ensure the requisite availability even in the event of a fault. The node SX1 now obtains, on client A's instructions, definite offers from various travel providers, which also contain, for example, the availability of specific dates, and offers them to client A for selection. In Fig. 5, the nodes SYl, SY2 and SY3 represent various travel providers. In this case, nodes which have not been incorporated in the virtual cluster NX are thus accessed. Steps 3a, 3b and 3c for enquiring with the travel providers SYl, SY2 and SY3 are independent of the selected node SX1 or SX2 since both nodes, owing to the formation of the cluster, may be regarded as being equal in terms of making the enquiry. Fig. 6 shows a further phase in the process of booking a trip. In a step 4, a trip is selected for definite and the booking is made. For this purpose, it is assumed that one of the three nodes SY1, SY2 or SY3 has the desired offer. In the example, it is assumed that this is the node SY1. Owing to the financial importance of booking a trip, the node SY1 must also have a high level of availability and therefore enters into a connection (needed to provide functions which increase availability) with a second node SY2. This is made possible by inventive coupling means at the nodes SY1 and SY2. SY1 has already previously determined SY2 as the backup computer (step 5). SY2 is therefore already in a high-availability relationship with respect to SYl and thus forms a virtual cluster NY with the latter. An advantageous trait is that SY2 may be identical to SX2. The same node is therefore part both of the virtual cluster NX and of the virtual cluster NY. This also constitutes an advantage over conventional local clusters since, in conventional clusters, it is not possible to incorporate a node or server in two clusters. This means that, in the event of a fault in SYl, a booking enquiry is directed from the node SX1 to the node SY2. This is shown in Fig. 6 by virtue of the connection for step 4 being between the two virtual clusters NX and NY, not between specific nodes in these clusters. Should SXl also fail in addition to SYl, client A's enquiry is answered solely by the node SX2 = SY2 since both the connection service to client A and the booking service which client A has selected run there. Finally, in the illustration in Fig. 7, the travel documents are prepared for the customer and the customer is billed in steps 6 and 7. If desired, online debiting may also be effected. In step 6, the suitable node for making the booking and for account settlement is first of all sought. Said node results, for example, from the bank which the customer has specified for automatic debiting. A node of this type is the node SZ1 in the illustration in Fig. 7. In contrast to the preceding steps, a configuration in which the service used is carried out in a completely different virtual cluster is shown here. It is, of course, important for high availability that the connection 6 is also redundant in order to make it possible to use a plurality of connection paths even in the event of a fault. This is also associated with the selection of a suitable partner node. It should also be noted that the network NX no longer has a direct relationship with respect to a network NZ which is formed by the node SZl and a second node SZ2. This is due to the fact that the selected travel service provider in the network NY is using the booking service provider in the network NZ. Furthermore, the travel provider has no control over, or information on, the network NZ. A plurality of steps may in turn be required within the virtual cluster NZ or in order to set up the latter. By way of example, the node NZ1 first of all tries to address a node SZa or a node SZb in order to form a virtual cluster with them. In a bank, however, it is conceivable that the node SZa will not be suitable for making bookings but rather will be intended only to provide customer information. The possibility of forming a cluster is therefore prevented by appropriately configuring the node SZa. Within the virtual cluster NZ, it is of course also possible to switch over between various nodes in the event of a fault if said nodes have corresponding high-availability functionality. In the instance of application just described, each link in the chain thus has a high level for availability for the purpose of carrying out a complex service. This is not achieved by means of individual nodes with clusters being equipped to have a high level of availability but rather by means of a flexible connection to other nodes, with the nodes each having inventive coupling means which provide functions which increase availability when two nodes are connected. In some cases, it may be advantageous to be able to transparently discern, from calling nodes or clients, which nodes are currently incorporated or active. An advantageous option is thus to translate physical addresses of nodes or data processing apparatuses into logical addresses. In this case, the physical addresses may be predefinable or automatically determinable. It is also advantageous if event and alarm messages are routed via the same channels and are centrally accessible, so that suitable countermeasures, for example termination of the service, may be initiated automatically or manually. It cannot always be assumed that there is a second node or a second data processing apparatus which can provide the same service and furthermore has the suitable inventive coupling means in order to enter into an availability-increasing connection with the first node, that is to say to form a virtual cluster. In this case, it proves to be advantageous in turn that the inventive coupling means can be implemented solely by means of software. It is therefore possible, on a node which, although it provides the service sought, does not have the inventive coupling means, to install precisely said coupling means remotely. This provides the greatest possible flexibility when setting up a virtual cluster. The instance of application described proceeds in an IP network, namely the Internet. However, the invention is not restricted thereto. The invention is applicable whenever a cluster needs to be formed in a network which has data processing apparatuses which are not identified. We Claim: 1. Data processing apparatus having a communication interface for connecting the data processing apparatus associated with a first network node to a network, characterized by coupling means for providing availability-increasing functions in conjunction with at least one further data processing apparatus which has coupling means, which is associated with a further network node and which is connected to the network, the network being a network of the kind in which data processing apparatuses are not registered and configurable at a central location. 2. Data processing apparatus as claimed in claim 1, wherein the coupling means are designed in such a way that they automatically provide the availability- increasing functions. 3. Data processing apparatus as claimed in Claim 1, wherein there are means for jointly administering the availability-increasing functions of the data processing apparatus and of the at least one further data processing apparatus. 4. Data processing apparatus as claimed in Claim 1, wherein the availability-increasing function of the coupling means in the data processing apparatus carries out memory replication from the data processing apparatus to the further data processing apparatus. 5. Data processing apparatus as claimed in Claim 1, wherein the availability-increasing function carry out a parallel calculation on the data processing apparatus and on the at least one further data processing apparatus. 6. Data processing apparatus as claimed in Claim 1, wherein the availability-increasing functions initiate translation of physical addresses of the least one further data processing apparatus, said physical addresses being predefinable or automatically determinable, into logical addresses, with the translation being carried out by the data processing apparatus. 7. Data processing apparatus as claimed in Claim 1, wherein the availability-increasing functions route event and alarm messages via uniform channels, with the messages being centrally accessible. 8. Data processing apparatus as claimed in Claim 1, wherein the availability-increasing function of the coupling means in the data processing apparatus provides timer objects of the at least one further data processing apparatus in redundant form. 9. Data processing apparatus as claimed in Claim 1, wherein the availability-increasing function of the coupling means in the data processing apparatus monitors the addressability of the at least one further data processing apparatus. 10. Data processing apparatus as claimed in Claim 9, wherein there are means for terminating availability-increasing functions by coupling to the at least one further data processing apparatus which can no longer be addressed. 11. Data processing apparatus as claimed in Claim 1, wherein the availability-increasing function of the coupling means in the date processing apparatus automatically installs an administration program unit from the first data processing apparatus on the at least one further data processing apparatus. 12. Method for increasing the availability of services on a data processing apparatus of a node in a network which is based on the Internet Protocol, having the following steps: a coupling means having availability-increasing functions is provided on the data processing apparatus, a request is made to a further data processing apparatus, associated with another node in the network, by the data processing apparatus regarding whether the further data processing apparatus is ready to provide services for increasing the availability of the data processing apparatus, the network being a network of the kind in which data processing apparatuses are not registered and configurable at a central location, and if the request in the last step is answered in the affirmative by the further data processing apparatus, then the two data processing apparatuses are coupled by the coupling means to form a virtual cluster and the availability-increasing functions of the coupling means are performed. 13. Method as claimed in Claim 12, wherein the coupling means is installed remotely on the further data processing apparatus if the further data processing apparatus does not have a coupling means. Data processing apparatus having a communication interface for connecting the data processing apparatus associated with a first network node to a network, characterized by coupling means for providing availability-increasing functions in conjunction with at least one further data processing apparatus which has coupling means, which is associated with a further network node and which is connected to the network, the network being a network of the kind in which data processing apparatuses are not registered and configurable at a central location.

Documents:

34-kolnp-2004-granted-abstract.pdf

34-kolnp-2004-granted-claims.pdf

34-kolnp-2004-granted-correspondence.pdf

34-kolnp-2004-granted-description (complete).pdf

34-kolnp-2004-granted-drawings.pdf

34-kolnp-2004-granted-examination report.pdf

34-kolnp-2004-granted-form 1.pdf

34-kolnp-2004-granted-form 18.pdf

34-kolnp-2004-granted-form 2.pdf

34-kolnp-2004-granted-form 26.pdf

34-kolnp-2004-granted-form 3.pdf

34-kolnp-2004-granted-form 5.pdf

34-kolnp-2004-granted-reply to examination report.pdf

34-kolnp-2004-granted-specification.pdf