面向公有云存储的高并发关键技术研究及系统实现

发布时间：2018-01-17 21:37

本文关键词：面向公有云存储的高并发关键技术研究及系统实现　出处：《华南理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：近年来云存储服务已逐渐成熟，许多企业和用户选择将数据存储在云上，其中面向互联网的云存储被称为公有云存储。公有云存储系统不仅要适应海量业务数据存储的增长，还要具备大规模、高并发处理访问的能力，以快速响应用户的大量请求。因此，如何有效提高公有云存储系统的并发处理能力是当前云存储服务领域的项研究重点，具有重要的实际意义。首先，本文介绍了两项能够提供并发能力的关键技术：负载均衡和反向代理缓存。以LVS为例深入研究负载均衡常用的算法和工作模式，，并简要地研究了反向代理缓存的实现原理。其次，深入分析研究OpenStack的对象存储Swift项目，建立了个基于负载均衡的多节点公有云存储模型，通过水平扩展代理节点数量即可线性提高系统的并发性能；同时，考虑到系统中存在高频访问的热点存储对象，文章还提出了个基于热点缓存的公有云存储模型，将访问热点的请求转发到反向代理缓存，从而降低了对存储几点的访问内压力，提高系统响应速度。然后，提出基于负载均衡的热点缓存公有云存储模型的高并发公有云存储系统框架，并详细设计出各个功能模块组成及其具体实现方法。其中，对原有Proxy Server加入了热点存储对象判断的功能改进，使得公有云存储系统能自动识别出获取热点存储对象的请求并将其转发交由缓存处理。经测试实验证明，本文提出的高并发公有云存储系统，在多用户访问场景下随着模拟并发数的增大，其并发处理能力呈线性上升趋势增长；而在模拟热点存储对象访问场景中，其TPS（Throughout Per Second）比单代理节点的公有云存储系统提升了8.48%，比多代理节点的公有云存储系统提升了4.29%。综上，基于提议模型搭建的原型系统切实有效地提高了公有云存储服务的并发性能。
[Abstract]:In recent years, cloud storage services have gradually matured, many enterprises and users choose to store data on the cloud. Internet oriented cloud storage is called public cloud storage. Public cloud storage system should not only adapt to the growth of massive business data storage, but also have the ability of large scale and high concurrent processing access. Therefore, how to effectively improve the concurrency processing capability of public cloud storage system is the focus of current cloud storage service research, and has important practical significance. First of all, this paper introduces two key technologies that can provide concurrent capability: load balancing and reverse proxy caching. Taking LVS as an example, the common algorithms and working modes of load balancing are studied in depth. The implementation principle of reverse proxy cache is briefly studied. Secondly, a multi-node public cloud storage model based on load balancing is established by deeply analyzing the object storage Swift project of OpenStack. The concurrency performance of the system can be improved linearly by horizontal expansion of the number of proxy nodes. At the same time, considering the existence of hot storage objects with high frequency access in the system, a public cloud storage model based on hot spot cache is proposed, which forwards requests for access to hot spots to reverse proxy cache. Thus, the internal pressure of access to several points of storage is reduced and the system response speed is improved. Then, a high concurrent public cloud storage system framework based on load balancing model of hot cache public cloud storage is proposed, and the composition of each function module and its specific implementation method are designed in detail. The original Proxy Server is improved by adding the function of judging the hot storage object. The public cloud storage system can automatically identify requests for hot storage objects and forward them to the cache. The test results show that the concurrency processing ability of the high concurrent public cloud storage system increases linearly with the increase of the number of simulation concurrency in the multi-user access scenario. In the simulated hot storage object access scenario, its TPS(Throughout Per second is 8.48% higher than the public cloud storage system of a single proxy node. Compared with the public cloud storage system with multi-agent nodes, the proposed prototype system improves the concurrent performance of public cloud storage service effectively and effectively.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333

【参考文献】