High-performance Fault Tolerance For Memcached Using Array Codes

With the rapid development of Internet technology,Web application traffic and its data amount are increasing sharply.Nowadays,most Web services store their data in relational database management systems(RDBMS).However,due to the low efficiency of RDBMS,database has already become the bottleneck of large-scale Web applications.In order to solve the above problems,some in-memory object caching systems have emerged in recent years,Memcached is a typical representative.As the cache,Memcached servers store data in memory and a repeated read operation can be processed without database which reduces the pressure of database access.Along with the high processing efficiency of memory,the performance bottleneck which is brought by database can be fundamentally solved.Now,Memcached is widely used and the distributed Memcached is the primary application because of the large data volume.As a distributed system,distributed Memcached should have the ability of fault-tolerance,that's,when a server node is crashed,the system should reconstruct the lost data to avoid the relevant operation pressure back to database.Unfortunately,the traditional distributed Memcached does not propose the corresponding solution for the above problem.To realize a fault-tolerant Memcached,related researches have been launched and some achievements have been made.For example,Cocytus uses Reed-Solomon codes(RS codes)and distributed protocols to realize the fault-tolerance mechanism.But the relatively complex finite field calculation used by RS codes and the high network transmission cost during data reconstruction are becoming a new performance bottleneck of Memcached.To solve these problems,this paper designs and implements a new distributed Memcached fault-tolerant framework.This framework simplifies the data reconstruction process and implements XOR-based array codes(RDP codes),which optimized calculation performance when reconstructing data.At the same time,in order to reduce the cost of network transmission during data reconstruction processes,this paper also optimizes the data reconstruction performance using an existing optimized scheme(RDOR),which leads to a high-performance fault-tolerance distributed Memcached.