Werner Schaad, Hans-J. Schek, and G. Weikum. Implementation and performance of multilevel transaction management in a multidatabase environment. In Proceedings of International Workshop on Research Issues in Data Engineering, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....this section by pointing out that all previously proposed checkpointing schemes, see in Table 4, exepts Gray s violate autonomy. 5.2. Use of Agents in Multidatabase Environment To circumvent the autonomy restriction an agent (or server) as seen in Figure 2, is often placed above the local DBMS [5, 14, 44]. Although the agent will typically perform other duties, we concentrate on those that would be performed to assure recoverability: 2PC, CC, logging. Note that an agent can not force an autonomous DBMS site to take a checkpoint or recover after a failure at another site. The primary function of ....

....we concentrate on those that would be performed to assure recoverability: 2PC, CC, logging. Note that an agent can not force an autonomous DBMS site to take a checkpoint or recover after a failure at another site. The primary function of the agent is to participate in a global automic commit [44, 60] To ensure recoverability of global transactions, the agent needs to maintain its own log which is in addition to that for the local DBMS [14] This log contains records which record the state of global subtransactions executing at the corresponding local DBMS. In addition, the log may contain ....

[Article contains additional citation context not shown here]

Werner Schaad, Hans-J. Schek, and G. Weikum. Implementation and performance of multilevel transaction management in a multidatabase environment. In Proceedings of International Workshop on Research Issues in Data Engineering, 1995.

....them up, we exploit intra transaction parallelism by breaking the updates into shorter relational operations. These are executed as ordinary independent parallel transactions on the relational storage server. This works builds on our previous experience with multi level transactions [Wei91, WS92a, SSW95] In particular, we explore and evaluate the key idea of exploiting inter transaction parallelism at the low level for intra transaction parallelism at the high level in the context of multi level transaction management as described in [WH93] Using parallelism to update derived data is also ....

....Finally, the Logging and Locking Components implement the level specific multi level transaction management. In this paper, we concentrate on the schema and update operation translators and on the logging component. The locking component is similar to the lock managers described in [Has95, SSW95] Details about the query translator, the paralleliser and the lock manager are beyond the scope of this paper. As we mentioned above, the reduction in retrieval costs resulting from our mapping is achieved at the expense of increased complexity of update operations due to the data replication ....

[Article contains additional citation context not shown here]

W. Schaad, H.--J. Schek, and G. Weikum. Implementation and Performance of Multi--level Transaction Management in a Multidatabase Environment. In Proceedings of the 5th Internat'l Workshop on Research Issues on Data Engineering: Distributed Object Management, Taipeh, Taiwan, 1995.

....take considerable time. Since all of the coordination process is performed within the bounds of a global transaction the system must be capable of dealing with these long transactions. We adopt a multi level transaction model to avoid the long term blocking of other transactions (see [SWS91, WS92, SSW95] for details of such transaction models) 6 Conclusions We propose an approach to the coordination of application systems through the maintenance of explicit inter system dependencies over application data. These dependencies have both a static and a dynamic part. The static part specifies ....

W. Schaad, H.-J. Schek, and G. Weikum. Implementation and Performance of Multilevel Transaction Management in a Multidatabase Environment. In 5th Intl. Workshop on Research Issues on Data Engineering: Distributed Object Management, RIDEDOM '95, Taiwan, March 1995.

....External object 1 External object 2 External object 3 CO OPN Prototype interface In procs. IEEE International conference on Systems, Man and Cybernetics SMC 96 , Beijing, China, Oct. 14 17, 1996. This principle is currently investigated by other researchers in the field of federated databases [22]. 6. PROPERTIES OF RESULTING PROTOTYPES We have shortly sketched the principles that are applied to incrementally build prototypes from abstract specifications. It seems that the results concern principally the domain of software engineering for concurrent systems, in particular the improvement ....

W. Schaad, H.-J. Scheck, G. Weikum, "Implementation and Performance of Multi-level Transaction Management in a Multidatabase Environment", 5th Int. Workshop on Research Issues in Data Engineering RIDE-DOM, Taiwan, March 1995.

....transactions that may commit or abort independently. Our system must guarantee atomicity on two levels: 1) atomicity of a single service request, and (2) atomicity of global transactions that comprise more than one service request. To this end, different techniques have already been applied: [22] implements atomicity by compensation at the 7 SQL level. That implies that for each insertion statement the log manager has to log the compensating SQL delete statement for each tuple inserted. With such an approach, logging yields a big overhead. In [17] the authors tackle this problem as ....

W. Schaad, H.-J. Schek, and G. Weikum. Implementation and performance of multi-level transaction management in multidatabase environment. In Proc. of RIDE-DOM'95 Taipei, Taiwan, pages 108--115, 1995. 25

....of the whole workflow is delegated to the more abstract workflow level where the information is available on what interleavings should not be allowed. As an example of the research efforts underway, BDS 93] has proposed to combine this technology with workflow specification. In [DSW94, SSW95] these ideas are used as an alternative for the 2PC protocol in distributed transaction processing. ffl Exception Handling: In workflow environments, the system cannot arbitrarily abort an operation. If an exception arises, human interaction is often required to resolve it, as it is generally ....

W. Schaad, H.-J. Schek, and G. Weikum. Implementationand Performance of Multi-Level Transaction Management in a Multidatabase Environment. In Proceedings of the 5th International Workshop on research Issues on Data Engineering, Taipei, Taiwan, March 1995.

....Fig. 2.2. Functionality necessary to support distributed process execution 2.2 Software Solutions Existing solutions to distributed execution are commonly grouped under the name of middleware. Middleware products related to databases are, for instance, federated and multi databases systems [SSW95, DSW94] TPmonitors [Obe94] persistent queuing systems [MH94] CORBA implementations [COR95a] workflow management systems [Hsu95, Hsu93] and process centered environments for software engineering [TKP94, BK94] They can support the process in the example in a variety of ways. For instance, a ....

.... This is best captured by using a nested transaction or multilevel model [Wei91, Mos81] which provides a powerful mechanism to reason about recovery in applications with a complex structure [GR93] A similar conclusion has been reached in several instances of middleware [BN97, Cor95c, CD96, SSW95] Thus, the navigator in OPERA is based on recent work that extends the existing notions of nested and multilevel transactions and applies them to composite systems [ABFS97] The goal is to exploit the parallelism inherent in processes, guaranteeing at the same time that the order of execution ....

W. Schaad, H.-J. Schek, and G. Weikum. Implementation and performance of multi-level transaction management in a multidatabase environment. In Proc. of the 5. Int. Workshop on Research Issues on Data Engineering, Distributed Object Management, Taipei, Taiwan, 1995.

....we assume that operations invoked different objects always commute. The matrix shows whether two operations on the same object commute or not. Note in the general case, e.g. when SQL operations are used, the conflict test must consider all input parameters of the operation invocations [DSW94, SSW95] read write read Gamma1 write Gamma1 read write read Gamma1 write Gamma1 Figure 1: State independent Commutativity Relation Example 1: Reconsider the read and write operations. The description of the operations can be restated as following: ffl ....

W. Schaad, H.-J. Schek, and G. Weikum. Implementation and performance of multi-level transaction management in a multidatabase environment. Proc. of the 5. Int. Workshop on Research Issues on Data Eng.: Distributed Object Management, Taipei, Taiwan, 1995.

....subsystem that executes the step while the correctness of the whole workflow is delegated to the more abstract workflow level where the information is available what interleavings should not be allowed. BDS 93] has proposed to combine this technology with workflow specification. In [DSW94, SSW95] these ideas are used as an alternative for the 2PC protocol in distributed transaction processing. ffl Exception Handling: In workflow environments, the system cannot arbitrarily abort an operation. If an exception arises, human interaction is often required to resolve it, as it is generally ....

W. Schaad, H.-J. Schek, and G. Weikum. Implementation and Performance of Multi-Level Transaction Management in a Multidatabase Environment. In Proceedings of the 5th International Workshop on research Issues on Data Engineering, Taipei, Taiwan, March 1995.

No context found.

Werner Schaad, Hans-Jorg Schek, and Gerhard Weikum. Implementation and performance of multi-level transaction management in a multdatabase environment. In Proceedings of the Workshop Research Issues in Data Engineering--Distributed Object Management (RIDE-DOM95), Taipei, Taiwan, March 1995.

No context found.

W. Schaad, H.-J. Schek, and G. Weikum. Implementation and performance of multi-level transaction management in multidatabase environment. In Proc. of the 5th Int. Workshop on Research Issues on Data Engineering: Distributed Object Management, RIDE-DOM'95, Taipei, Taiwan, Mar. 1995. inproceedings.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC