Research On Performance And Power Consumption Optimization Of Distributed Cache System Based On Consistent Hash

In recent years,network data has shown a multi-channel,multi-category and multi-way growth trend,the popularity of the concept of big data reflects the importance of network data in today's society as well.Faced with the blowout growth of data,how to access data quickly and efficiently has become the focus of researchers' attention,and the application of distributed data caching system makes data storage have a new solution.This paper mainly studies the Memcached distributed caching system,and proposes effective solutions to the heterogeneity of system servers and the high-power consumption of the system.These methods optimize and improve the cache performance and power consumption of the system.Aiming at the heterogeneity of system servers,on the one hand,this paper proposes a gradient consistent hash data caching method.This method divides the servers with similar performance into a gradient,and configures the gradients proportionally to reduce the amount of virtual servers and improve the adaptability of the server cluster.On the other hand,the gradient load balancing strategy is used to optimize the server load,which improves the overall performance of the system.Aiming at the problem of high-power consumption,a multistage power replica layout strategy based on consistent hashing,Multi_PowerST,is proposed.The strategy divides the servers into different gradients,and sets three different power consumption modes for the system according to the proportion of open servers:low(open 30%),medium(open 60%)and high(open 100%).By maintaining the working state of different power consumption modes,servers in other power consumption modes can be put into standby state without affecting data caching.Thus achieving the goal of saving power consumption.The simulation results show that the gradient consistent hash data caching method proposed in this paper can improve the overall performance of the server to a certain extent and optimize its load balancing problem.The Multi_PowerST strategy proposed in this paper is more energy-saving than the original Memcached system.