The Improvement Research Of Distributed Caching System Memcached

The continuous development of the Internet companies brought not only convenience to people's online life experience,but also a large increase of the amount of data.As a result,the storage and high-speed access of a large amount of data was now one of the biggest challenges faced by the age of the Internet development.The concept of cache introduced to improve the overall performance of the Web application system by dealing with read/write speed between CPU and memory does not match.In order to avoid this situation that a centralized cache server lead to system performance degradation by swap cache data frequently,web application system introduced distributed cache system.Distributed cache system had high availability,good scalability and the characteristic of the structure is transparent to users.Its related research began in the 1990s,lots of studies had already been applied to various large Web network applications.The problem of concentrated or severe inclined access load can make the single node of distributed cache system appeared big pressure problem.Firstly,this paper introduces and analyzes the distributed cache technology,including distributed topological structure,data caching strategies and typical replacement strategies,then it introduces the technology of the distributed cache system proposes in this paper.Secondly,this paper briefly introduces characteristics and scenarios of Memcached and Redis which are typical distributed caching system.Memcached and Redis are widely used in the internet these days.Under aiming at centralized access load scenarios,this article will improve on distributed cache system Memcached.This paper detailed researches and analysis distributed cache system Memcached's client-side and server-side,so as to summarizes and analysis the shortcomings of its.Because original Memcached system do not have redundancy mechanisms,cache server may have performance problems and single-point failure problem if there is a sudden situation.This paper designs and introduces the MS_Memcached distributed cache system which add master-slave mode to Memcached,and the MS_Memcached introduces the problems of master-slave consistency,requests "vertical" load balancing and single-point failure.The MS_Memcached realizes master-slave replication through the way of asynchronous message queue in this paper,realizes requests "vertical" load balancing through dynamic feedback load balancing algorithm,solves the problem of single-point failure by using the election algorithm.Furthermore,MS_Memcached distributed cache system introduces Amazon's data partitioning scheme to Memcached for more uniform.In the experimental part,this paper builds a Memcached cache server cluster for the experiment.The cluster is comprised of one PC and six cache server machines.The experiment use Memcached Java client whalin₂.5.0_bin official test data.According to the experimental results,on the basis of the performance has not been affected by the scheme proposed in this paper,in the case of appear access load tilt,the number of requests per second in the master-slave MS_Memcached distributed caching scheme which is designed in this paper increased nearly 60%than that in original Memcached scheme under the condition of multi thread.The master-slave MS_Memcached distributed caching scheme can better solve the problem of serious tilt access load.Under the scenario of a single point of failure,in this paper,the master-slave MS_Memcached architecture also has a better ability of troubleshooting than original Memcached scheme,and MS_Memcached scheme is failures-transparent to users.By improving the Hash consistency virtual node distribution scheme of original Memcached,the MS_Memcached scheme makes the distribution of data on each server approach to the ideal.