《计算机毕业设计外文翻译》由会员分享，可在线阅读，更多相关《计算机毕业设计外文翻译（12页珍藏版）》请在金锄头文库上搜索。
1、毕业设计（论文）外文资料原文及译文专 业 班 级 学 号 姓 名 指 导 教 师 毕业设计（论文） 外文资料 原文1原文出处： by Michael J. Carey/Michael J. Franklin/Miron Livny/Eugene J. Shekita ,Computer Sciences Technical Report #994, January 1991Data Caching Tradeoffs in Client-Server DBMS ArchitecturesIn this paper, we examine the performance tradeoffs th
2、at are raised by caching data in the client workstations of a client-server DBMS. We begin by presenting a range of lock-based of the well-understood problem of replicated data management. We then use a detailed simulation model to study the performance of these algorithms over a wide range of workl
3、oads and systems resource configurations. The results illustrate key performance tradeoffs related to client-server cache consistency, and should be of use to designers of next-generation DBMS prototypes and products.1. INTRODUCTIONWith networks of powerful workstations becoming commonplace in scien
4、tific, engineering, and even office computing environments, client-server software architectures have become a common approach to providing access to shared services and resources. Most commercial relational database management systems today are based on client-server architectures, with SQL queries
5、 and their results serving as the basis for client-server interactions. In the past few years, a number of object-oriented DBMS prototypes and products have appeared, virtually all of which are based on client-server architectures. Compared to relational database systems, these systems generally tak
6、e a different approach when it comes to client-server interactions. In order to offload processing to client workstations, it is common for client-server interaction in an OODBMS to take place at the level of individual objects or pages of objects rather than queries. Prototypes based on object-leve
7、l interaction include Orion and O2, among others. Among the prototypes based on page-level interactions are ObServer and the current version of the EXODUS storage manager; the ObjectStore system from Object Design is an example of a commercial OODBMS product based on page-level interactions.In archi
8、tectures where data pages or objects are the basis for client-server interactions, it is possible to cache data in the local memories of client workstations for later reuse. Such caching can reduce the need for client-server interaction, lessening the network traffic and message processing overhead
9、for both the server and its clients. It also enables client resources to be used by the DBMS, thus increasing both the aggregate memory and the aggregate CPU power available for database-related processing. An increase in the aggregate memory of the DBMS can reduce the I/O load on the server, while
10、an increase in the aggregate CPU power available to the DBMS can reduce the load on the server CPU(s).Depending on the nature of the applications-including their balance of I/O and CPU demands, their locality of access, and the proportional cost of their DBMS accesses relative to their overall compu
11、tational requirements-increasing the 毕业设计（论文） 外文资料 原文2aggregate resources of the DBMS can result in significant performance improvements. Of course, in applications where the response time is dominated by the time spent at the client CPU, or where a large fraction of the DBMS is accessed relative to
12、 the size of the client buffer pool, the performance benefits of caching would be negligible.Despite its potential, caching is not a performance panacea. In order to incorporate caching, the DBMS must include a protocol that ensures cache consistency. Such a protocol may be complex to implement, it
13、may entail a significant amount of processing overhead, and its impact on system performance may be workload-dependent. Depending on how the protocol and the workload interact, the cache consistency protocol might actually increase the load on the server and/or the client workstations due to its ove
14、rhead, particularly when there are a large number of client workstations. Another potential pitfall, which depends on the concurrency control scheme used by the protocol, is the late discovery of data conflicts. Thus, the potential consequences of adding caching to a client-server DBMS range from a
15、significant improvement in performance to a notable degradation in performance. An example of a workload where caching can be highly beneficial is the Sun Engineering Database Benchmark.In this paper, we examine the data caching performance tradeoffs discussed above. We begin by presenting a range o
16、f lock-based cache consistency algorithms that result from recognizing that cache consistency is simply a variant of the replicated data management problem studied by distributed DBMS researchers. For concreteness, we focus our attention on systems where client-server interaction are page-based. This approach, also referred to as the block server approach, was shown to perform well for CAD-style data access patterns in a recent performance
