In the client-server database systems, clients can maintain their own cache to reduce the overhead of accessing data objects at server databases. In this case: there should be a protocol which ensures the consistency of cahed data, and it can be divded into two classes: pessimistic protocol and optimistic protocol.In this paper, we propose a new cache management scheme which dyuumically applies an appropriate protocol for each data object based on the degree of contention on each of them.When the contention of data objects is uniformly high or uniformly low, the proposed scheme shows the same performance with conventional algorithms, and when the contention is not uniform, the proposed scheme shows better performance

No abstract

Abstract:Data management for distributed computing has spawned a variety of research work and commercial products. At the same time, recent technical advances in the development of portable computing devices and the rapidly expanding cordless technologies have made the mobile computing a reality. In conjunction with the existing computing infrastructure, data management for mobile computing gives rise to significant challenges and performance opportunities. Most mobile technologies physically support broadcast to all mobile users inside a cell. In mobile client-server models, a server can take advantage of this characteristics to broadcast information to all mobile clients in its cell. This fact introduces new mechanisms of data management which are different from the traditional algorithms proposed for distributed database systems. In this chapter, we give executive summary and discuss topics such as data dissemination techniques, transaction models and caching strategies that utilize broadcasting medium for data management. There is a wide range of options for the design of model and algorithms for mobile client-server database systems. We present taxonomies that categorize algorithms proposed under each topic. Those taxonomies provide insights into the tradeoffs inherent in each field of data management in mobile computing environments.

Load balancing is one of the best efficient methods for performance improvement of cluster system. Recently, WLC algorithm is used for the load balancing of cluster system. But, the algorithm also has load imbalance between servers, because it uses inaccurate static load status of servers. In this paper, I suggest a more efficient dynamic load balancing algorithm base on various load status information of servers by real time. It shows that load imbalance phenomenon is improved greatly and response time is also improved compare with WLC algorithm.

No abstract