Research And Implementation Of Cache System In Distributed Full-Text Searching

Cache technology is a pivotal way to improve the respond speed and system efficiency. It can promote related properties of system in many fields. In distributed full-text searching process, cache technology has a positive value. The research of cache technology of distributed full-text searching has both high theoretical value and commercial basis. On public opinion monitoring system what is a national important project, through the analysis of users'query results, this task designed a distributed cache system of query logs. From the test of experiment, the cache system has better results on hit rate, and improves processing speed of query.According to the needs of public opinion monitoring system, design architecture of cache system for distributed full-text searching, finish all the functions of the distributed caching system, including system architecture, cooperation mode, cache structure, caching replacement algorithm and the other aspects.This paper designs the two-level cache structure that has not only the dynamic cache, but also has the static cache. Through statistical analysis of query logs, the hottest queries are extracted to store in the static cache. When queries arrive, firstly, cache system searches the static cache, according to the fact of cache hit, then decides whether need to search the dynamic cache or index. For the static cache contains the records of the hottest queries, with the assistance of dynamic cache, cache system with this cache structure has more cache hits and faster processing speed. Comparing with the other cache structure, the two-level cache structure has a certain advantage.And, by analyzing the traditional cache replacement algorithm, based on the theoretical basis of LRU and LFU. According to the features of cache structure, design a caching replacement algorithm. The algorithm deals not only the replacement of dynamic cache, but also the replacement between static cache and dynamic cache. Comparing with the traditional caching replacement algorithm, it gets better hit rate.