| Today, information exchange has entered the era of massive data. As an important measure to enhance data access speed, distributed cache technology has received widespread attention from industry and academia in recent years. Memcached is the most famous open-source distributed cache product, which plays an extremely important role in cache areas, therefore, research and optimization for its performance is particularly necessary. The purpose of this paper is to promote memory utilization and cache hit rate of memcached in the case of guaranteed throughput. This research has great scientific research and engineer apply value, and provides an economical and practical solution for Internet services.Based on the research status of memcached, this paper analyzed the current research progress about its internal optimization. And for the current weakness of domestic research, this paper established a research idea to enhance memory efficiency by improving the method of data expired. Focused on several existing cache replacement algorithm, and against the limitation of LRU algorithm and the lack of LIRS algorithm, this paper presented a replacement strategy. The new algorithm was more adaptable to distributed cache system, named hierarchy cache replacement algorithm based on LIRS. Simulation of several algorithms above was completed by C++ language, and comparison was analyzed in detail.In the design and implementation level, the new algorithm combined with lazy expiration has applied to memcached instead of the original LRU algorithm. With the data set of Zipf distribution and Mixed TTL, the performance of two systems was tested under different scenarios. Through the comparison we can see the improved system was better, which has improved memory space utilization while maintaining throughput and latency, reduced cache miss rate to half of original system, and made system cache hit rate increase by 20%. All in all, the optimization effect of memory utilization was obvious. |
PDF Full Text Request