A Server Caching Strategy in Client-Server Database Systems

this paper, we propose a server caching strategy which take advantage of the replaced page-id from the clients as the major information. Since the replaced page-id is only sent to the server by piggybacking whenever a new page fetch request is sent, the operation to deliver the replaced page-id is simple and induces a minimal overhead. Once the replaced page-id arrives on the server: 1) it guarantees that the corresponding page does not exist in the client cache, 2) it is expected to have highest reference probability among the pages in the server cache, and 3) it is used as an additional information besides page. Consequently, it alters the location of the corresponding page in the server cache, which means adjusting the importance of the page in the server. We show the proposed strategy reveals good performance characteristics on diverse situations such as single and multiple clients as well as various access patterns. The performance of our proposed algorithm is 110% higher than that of LRU and 34% higher than that of 2Q[4] in case of a single client system. As far as we know, this approach is the first attempt in this area, and also differs from the Franklin's work[3]. The latter focused on utilizing clients' buffer pool, and used the page-id of the page to be replaced to decide whether the server should copy back the page from the client. Meanwhile our proposed strategy uses the replaced page-id (id of the replaced page) as the most important information. Willick[8] recommended the FBR algorithm[7] for the server caching. However, as shown in several experiments in Section 4, the FBR suffers from a considerably low hit-ratio for the varying reference pattern. 2. Properties of LRU page replacement mechanism Many studies for the LRU algorithm focused on the hit-ratio..