Electrical computers and digital processing systems: multicomput – Computer-to-computer data routing – Least weight routing
Reexamination Certificate
1999-07-12
2001-10-09
Vu, Viet D. (Department: 2154)
Electrical computers and digital processing systems: multicomput
Computer-to-computer data routing
Least weight routing
C709S203000, C709S223000, C709S241000, C707S793000
Reexamination Certificate
active
06301601
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to a server application programming model using software components, and more particularly relates to the component having control over whether a transaction in which the component participates can be committed.
BACKGROUND AND SUMMARY OF THE INVENTION
In many information processing applications, a server application running on a host or server computer in a distributed network provides processing services or functions for client applications running on terminal or workstation computers of the network which are operated by a multitude of users. Common examples of such server applications include software for processing class registrations at a university, travel reservations, money transfers and other services at a bank, and sales at a business. In these examples, the processing services provided by the server application may update databases of class schedules, hotel reservations, account balances, order shipments, payments, or inventory for actions initiated by the individual users at their respective stations.
Often, server applications require coordinating activities on multiple computers, by separate processes on one computer, and even within a single process. For example, a money transfer operation in a banking application may involve updates to account information held in separate databases that reside on separate computers. Desirably, groups of activities that form parts of an operation are coordinated so as to take effect as a single indivisible unit of work, commonly referred to as a transaction. In many applications, performing sets of activities as a transaction becomes a business necessity. For example, if only one account is updated in a money transfer operation due to a system failure, the bank in effect creates or loses money for a customer.
A transaction is a collection of actions that conform to a set of properties (referred to as the “ACID” properties) which include atomicity, consistency, isolation, and durability. Atomicity means that all activities in a transaction either take effect together as a unit, or all fail. Consistency means that after a transaction executes, the system is left in a stable or correct state (i.e., if giving effect to the activities in a transaction would not result in a correct stable state, the system is returned to its initial pre-transaction state). Isolation means the transaction is not affected by any other concurrently executing transactions (accesses by transactions to shared resources are serialized, and changes to shared resources are not visible outside the transaction until the transaction completes). Durability means that the effects of a transaction are permanent and survive system failures. For additional background information on transaction processing, see, inter alia, Jim Gray and Andreas Reuter,
Transaction Processing Concepts and Techniques
, Morgan Kauffmann, 1993.
In many current systems, services or extensions of an operating system referred to as a transaction manager or transaction processing (TP) monitor implement transactions. A transaction is initiated by a client program, such as in a call to a “begin
13
transaction” application programming interface (API) of the transaction monitor. Thereafter, the client initiates activities of a server application or applications, which are performed under control of the TP monitor. The client ends the transaction by calling either a “commit_transaction” or “abort_transaction” API of the TP monitor. On receiving the “commit_transaction” API call, the TP monitor commits the work accomplished by the various server application activities in the transaction, such as by effecting updates to databases and other shared resources. Otherwise, a call to the “abort_transaction” API causes the TP monitor to “roll back” all work in the transaction, returning the system to its pre-transaction state.
In systems where transactions involve activities of server applications on multiple computers, a two-phase commit protocol often is used. In general, the two-phase commit protocol centralizes the decision to commit, but gives a right of veto to each participant in the transaction. In a typical implementation, a commit manager node (also known as a root node or transaction coordinator) has centralized control of the decision to commit, which may for example be the TP monitor on the client's computer. Other participants in the transaction, such as TP monitors on computers where a server application performs part of the work in a transaction, are referred to as subordinate nodes. In a first phase of commit, the commit manager node sends “prepare_to_commit” commands to all subordinate nodes. In response, the subordinate nodes perform their portion of the work in a transaction and return “ready_to_commit” messages to the commit manager node. When all subordinate nodes return ready_to_commit messages to the commit manager node, the commit manager node starts the second phase of commit. In this second phase, the commit manager node logs or records the decision to commit in durable storage, and then orders all the subordinate nodes to commit their work making the results of their work durable. On committing their individual portions of the work, the subordinate nodes send confirmation messages to the commit manager node. When all subordinate nodes confirm committing their work, the commit manager node reports to the client that the transaction was completed successfully. On the other hand, if any subordinate node returns a refusal to commit during the first phase, the commit manager node orders all other subordinate nodes to roll back their work, aborting the transaction. Also, if any subordinate node fails in the second phase, the uncommitted work is maintained in durable storage and finally committed during failure recovery.
In transaction processing, it is critical that the client does not pre-maturely commit a server application's work in a transaction (such as database updates). For example, where the activity of a server application in a transaction is to generate a sales order entry, the server application may impose a requirement that a valid sales order entry have an order header (with customer identifying information filled in) and at least one order item. The client therefore should not commit a transaction in which the server application generates a sales order before both an order header and at least one order item in the sales order has been generated. Such application-specific requirements exist for a large variety of server application activities in transactions.
Historically, pre-mature client committal of a server application's work in a transaction generally was avoided in two ways. First, the server application can be programmed such that its work in a transaction is never left in an incomplete state when returning from a client's call. For example, the server application may implement its sales order generation code in a single procedure which either generates a valid sales order complete with both an order header and at least one order item, or returns a failure code to cause the client to abort the transaction. The server application's work thus is never left in an incomplete state between calls from the client, and would not be committed prematurely if the client committed the transaction between calls to the server application.
Second, the client and server application typically were developed together by one programmer or a group of programmers in a single company. Consequently, since the programmers of the server application and client were known to each other, the server application programmers could ensure that the client was not programmed to commit a transaction between calls that might leave the server application's work in an incomplete state. For example, the client's programmers could be told by the server application programmers not to call the commit_transaction API after a call to the server application that sets up an order header and before a call to the server appl
Al-Ghosein Mohsen
Helland Patrick James
Limprecht Rodney
Klarquist Sparkman Campbell & Leigh & Whinston, LLP
Microsoft Corporation
Vu Viet D.
LandOfFree
Disabling and enabling transaction committal in... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Disabling and enabling transaction committal in..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Disabling and enabling transaction committal in... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2574634