T. Böhme and H. Unger (Eds.): IICS 2001, LNCS 2060, pp. 116–125, 2001.  Springer-Verlag Berlin Heidelberg 2001 A General Adaptive Cache Coherency-Replacement Scheme for Distributed Systems Jose Aguilar 1 and Ernst Leiss 2 1 CEMISID, Dpto. de Computacion, Facultad de Ingenieria, Universidad de Los Andes, Merida 5101, Venezuela aguilar@ing.ula.ve 2 Department of Computer Science, University of Houston, Houston, TX 77204-3475, USA coscel@cs.uh.edu Abstract. We propose an adaptive cache coherence-replacement scheme for distributed systems that is based on several criteria about the system and appli- cations, with the objective of optimizing the distributed cache system perform- ance. We examine different distributed platforms (shared memory systems, dis- tributed memory systems, and web proxy cache systems) and the potential of in- corporating coherency-replacement issues in the cache memory management system. Our coherence-replacement scheme assigns a replacement priority value to each cache block according to a set of criteria to decide which block to re- move. The goal is to provide an effective utilization of the distributed cache memory and a good application performance 1 Introduction The performance of distributed caching mechanisms is an active area of research [2, 3, 4, 5, 7, 8, 10]. A distributed cache memory is the simplest cost-effective way to achieve a high-speed memory hierarchy. A cache provides, with high probability, instructions and data needed by the local CPU at a rate that is more in line with the local CPU’s demand rate. Many studies have examined policies for cache replacement and cache coherence; however, these studies have rarely taken into account the com- bined effects of policies [2, 5]. In this paper we propose an adaptive cache coherence- replacement scheme for distributed systems. Our approach combines classical coher- ence protocols (write-update and write-invalid protocols) and replacement policies (LRU, LFU, etc.) to optimize the overall performance (based on criteria such as net- work traffic, application execution time, data consistence, etc.). This work is based on previous work we have done on cache replacement mechanisms which have shown that adaptive cache replacement policies improve the performance of computing sys- tems [1]. The cache coherence mechanism is responsible for determining whether a copy in the distributed cache system is stale or valid. At the same time, it must update the invalid copies when a given site requires a block. We study the impact of our scheme in different distributed systems: shared-memory systems, distributed-memory systems, and web proxy systems.