اطمینان از ثبات معامله خدمات وب

Composite Web services provide promising prospects for conducting cross-organizational Web service transactions. Such transactions generally require longer processing time and manipulate financially critical data. To efficiently manage these Web services transactions, isolation is commonly relaxed, but inconsistency will be caused by concurrently executing isolation-relaxing transactions. This article proposes an extension to the WS-Business Activity Protocol, which ensures the consistent executions of isolation-relaxing Web service transactions, and which is based on transaction dependency graphs distributed over multiple nodes. Furthermore, this article presents several algorithms implementing the protocol, and introduces the implement of a prototype system.

Many IT organizations have been implementing interfaces for business partners using the Web services (WS) technology.Due to its flexibility, WS allows organizations to composite services across other organizations as well as within organizations. Composite Web services provide promising prospects for conducting cross-organizational Web service transactions. Such transactions generally require longer processing time and manipulate financially critical data. However, traditional protocols like the two-phase locking protocol and the two-phase commit protocol, are inappropriate for long running transactions. This is due to the fact that these protocols are based on the strict isolation policy which requires that a transaction should not share its data with other transactions until it is completed. Thus, if traditional protocols are used, a transaction may not be able to access its resources even for a few days, until other transactions release their resources. To efficiently manage these Web services transactions, isolation is commonly relaxed. But the isolation relaxation introduces a serious inconsistency problem. Considering a situation where several transactions (T1andT2).oncurrently invoke WS from different providers, one participant P1 inT1 needs to be compensated after it has updated its data item X, but another participant P 2 in T2 has already read X before P1 is compensated and X is changed back to its original value. In that case, the participant may be in conflict with another one from another transaction. Such a situation may occur quite frequently because WS transactions will get more and more prevalent [1]. Although most transaction models and protocols for WS transactions have been proposed, no sophisticated concurrency control mechanism exists. Thus, organizations must take the responsibility and make great efforts to solve local consistency and inter-transaction consistency.

The authors propose a dependency management protocol called WSTDCP. The protocol is based on transaction dependency graphs distributed over multiple nodes, and guarantees the consistent executions of isolation-relaxing WS transactions. In the protocol, the WS developers may be responsible for detection of ended-state dependencies because they know the data semantics best, and know when ended-state dependencies may happen better than anyone else. In the article, the authors present several algorithms implementing the protocol, and implement a prototype system to validate our protocol.