Key Technology Research Of High Concurrency And System Realization Oriented To Public Cloud Storage

As the cloud storage services become intelligent and customer-friendly these days, largequantities of companies and individuals are willing to store data in cloud, this technologyproviding services via the Internet is called public cloud storage system. Public cloudstorage system not only should meet the growth of mass data and files stored, but alsoshould response immediately to a large number of users’ requests by means of scaling andelevating the concurrency. Therefore, public cloud storage system has significant practicalmeaning to find an efficient way to increase the concurrency without limitation and enhancethe capability to deal with more requests simultaneously.Firstly, this paper introduces load balancing and reverse proxy cache to us, both of whichare the keys to elevate the system concurrency. Taking LVS for example, the paper hasstudied in-depth to the common load balancing algorithms as well as the operating modes.Then it has a brief glance at the implementation principle of the reverse proxy cache.Secondly, while deeply analyzing to the Openstack Object Storage project, called Swift, thispaper has established a multi-node public cloud storage model which is based on loadbalancing. It enables the system to increase concurrency linearly by extending the proxynodes horizontally. Taking hotspots into consideration, this paper also presents ahotspots-cached public cloud storage model, which forwards the requests of getting hotspotsto reverse proxy caches in order to decrease the pressure of accessing the storage nodes andspeed up the responses at the same time.Thirdly, based on the load balancing hotspots-cached public cloud storage model, the paperproposes a system framework of high concurrent public cloud storage, and details thedesigns and implementation of all modules. One of the most important functionalimprovements is that the public cloud storage system can recognize hotspots and forward the request to caches automatically by adding judgment in the Proxy Server.Finally, the experiments prove that through simulation, the proposed high concurrent publiccloud storage system,which proposes in this paper, can elevate the concurreny growthlinearly while dealing with the increasing large amounts of users’ accessing requests. On theother side, while dealing with a large quantity of requests for the same hotspot, the system sthroughout(TPS) is8.48%greater than the public cloud storage system with single proxynode, and4.29%greater than the public cloud storage system with multi-proxy nodes. Todraw the conclusion, based on the proposed model, the prototype system do improve theperformance of high concurrency of public cloud storage system effectively and efficiently.