分布式缓存系统的内存利用率和并发性能优化机制-计算机科学与技术专业论文.docx

华 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract Memcached is a high-performance, distributed memory object caching system. It caches objects in memory to reduce the number of database reads and improve theread and write speeds. With the rapid development of web applications, Memcached is required ofmore efficiency and concurrency performance and is playing an increasingly important role in the field of cache.Then an urgent problem occursof seeking an optimization mechanismto meet the needs of application services and performance. Theresult shows thedifficultiesof balancing memory efficiency and high concurrent requests using small object key storage and read-oriented workloads by analyzingthe characteristic ofMemcached workloads and the comparisonbetween different systems in memory efficiency and concurrency mechanism. This papperprovides the correspondingsolution named optimization of cuckoo hashing algorithm to replace thedefault system hash table. It can enhance the efficiency of Memcached memorywith high occupancy hash table, achieve a single writer multiple readers concurrent access to hash table with optimized hash table queryinsertion algorithm and maintain the operation of hash table cache friendliness with tagbasedquery and insertion. The optimistic concurrency controlleruses counter to replace the original system global lock and uses changeof version number to find concurrent replacement operation, then it can reduceinternal thread synchronization, ensure data consistency,support a large number of concurrent access and reduce insert query overhead. We achieved a memory efficiency and concurrency performance optimization in Memcached basing on cup-intensive computing with linux operating platform.The optimization improves multi readers single writer high concurrent access rateswhile maintaining a high efficiency of the memory, decreases the memory usage by 20% with same test data setwhich meets zipfan distribution and increases the throughput by at least two-fold