Patent Application
20090049173
The invention relates to computer programming using proxy servers for client/server network management.
The client/server model of distributed computing operates to fulfill user needs by splitting functions between “client” tasks and “server” tasks performed by various computer hardware and software resources that are organized into a “network” for communication with each other, such as a local area network (“LAN”) or a wide area network (“WAN”) or the Internet. Using this model, a “client” program sends message requests to a “server” program in order to obtain data and/or processing action according to some communication “protocol” (i.e., a set of standard rules that determine how information is transmitted across a network) and the server completes the processing transaction by carrying out the request or deferring it to another time or by indicating that it cannot be fulfilled. This model allows clients and servers to be located (and to operate) independently of each other in a computer network, often using different hardware and operating systems appropriate to the function of each.
A “proxy server (or gateway)” is often used in handling client requests for transactions to be completed by other network “application servers” which are capable of performing the data processing actions required for the transaction but are not accessed directly by the client. If a processing transaction is not successfully completed upon initial transmission of a message, the client may send retransmissions of the message to an application server using an “arrayed cluster” (or group) of proxy servers. In that case, the cluster of proxy servers must route the retransmission(s) to the same application server as the original transmission, so that a retransmission is identical to the original transmission using the standards of RFC 3261. Current proxy server technology (such as that used with IBM WebSphere®) provides a partial solution to this problem by addressing (i.e., “hashing) the message to an array of application servers. This solution will work as long as the number (and the relative processing load(s) or “weight”) of each of the clustered application servers does not change. However, a retransmitted proxy message may be routed to a different application server than originally intended if a server starts (or stops) functioning between retransmissions of a message (or if the “weight” of one of the servers changes).
The invention provides for dynamic client/server network management using proxy servers. Specifically, a programming method and system is used for allowing each active proxy server in an arrayed cluster to maintain an updated list of all other operating proxy servers in the cluster (referred to as a “ProxyClusterArray”). When a client message (requesting access to a networked application server) is received by a clustered proxy server, the message may be forwarded to another proxy server (within the cluster) so that message (re)transmissions can pass through the same proxy server as the original message, allowing a proxy server to make consistent routing (and other) decisions pertaining to that message.
When a proxy server receives a message from a user client requesting access to an application server in order to carry out a processing transaction, the proxy server “hashes” the message (by addressing it) to a “destination” proxy server. If the destination proxy server is not the one that initially received the message then it is forwarded to the original (destination) proxy server, which (locally) maintains processing “state” information for a sufficient period of time to determine if the message is a retransmission (and if so the destination proxy server can make the same processing decisions as were made for the original message). For example, the destination proxy server can identify (or “remember”) the application server to which the original message was addressed (or “routed”) in order to route retransmitted message(s) to the same application server. Each proxy server maintains such status information to indicate the processing decisions made for all messages it has handled (including identification of the application server to which a message has been routed) for a designated (i.e., maximum possible) interval of time between message retransmissions (referred to as MAX_STATE) which is adjusted to account for potential network transmission delays. This approach allows a proxy server to make consistent decisions for message (re)transmissions in a dynamically changing client/server network processing environment.
The present invention provides for dynamic client/server network management using proxy servers, by allowing each active proxy server in an arrayed cluster to maintain an updated list of all other operating proxy servers in the cluster.
The present invention provides a method and system that allows a user client to send a message requesting access to an application server via a clustered proxy server by using a proxy server to provide a message hash identifying the proxy server maintaining state information for that message.
The subject matter which is regarded as the invention is particularly pointed out and distinctly claimed in the concluding portion of the specification. The invention, however, together with further objects and advantages thereof, may best be understood by reference to the following description taken in conjunction with the accompanying drawings.
As shown in the flowchart of
Whenever a new proxy server 33 is activated (or deactivated) a new ProxyClusterArray is activated (and becomes the “latest”) and the (previously latest) array is then considered to “expire” (after the next MAX_STATE interval). Thus a message may be forwarded more than once between proxy servers for examination to determine if it is being retransmitted. However, in the normal case there is only a single listed server in the ProxyClusterArray to consider, and (N−1)/N (where N is the number of proxy servers) of the client messages will be forwarded once to another proxy server. In a preferred example of use of the invention with WebSphere, an “HAGroup” (High Availability Group) processing mechanism (based on “Virtual Synchrony” technology) is used to track the list(s) of active and expiring ProxyClusterArrays. When activated, each proxy server joins an HAGroup that corresponds to the cluster of which it is a member, so that Virtual Synchrony can ensure that each active proxy server is provided with the same updated list(s) of ProxyClusterArrays. If a proxy server purposefully stops its processing activity, it broadcasts its state information to the other proxy servers listed in the (currently active) ProxyClusterArray; however, if a proxy server terminates operation abnormally, the state information it possesses is lost. (It is possible to handle such a condition by replicating this state information, although this entails a processing performance cost.)
While certain preferred features of the invention have been shown by way of illustration, many modifications and changes can be made that fall within the true spirit of the invention as embodied in the following claims, which are to be interpreted as broadly as the law permits to cover the full scope of the invention, including all equivalents thereto.
