基于多存储介质的在线社交网络数据节能存储研究

Large social networking sites have deployed their services at multiple data centers around the world, in order to deal with widely distributed users as well as huge amount of data. Nowadays, we are advocating energy conservation and green computing. The energy consumption problem of large-scale social network storage system has begun to attract attention in both academia and industry. Existing studies, on the one hand, have not been fully integrated with the features of online social networks, and mainly focused on the optimization of both access performance and operating costs, while few studies took energy consumption into account; on the other hand, the solid-state drives have many advantages including high performance and low power consumption, but few studies considered the positive effects of such new storage media on energy optimization. To solve the above problems, this project intends to start from two aspects: physical resources and data resources. By considering energy consumption, cost and performance, we will combine both the service execution logic and interaction features of online social networks together, and start an in-depth research on energy efficient resource provisioning, energy efficient data distribution among multiple data centers, and energy efficient data scheduling based on multiple storage media. The objective of project is to make social network storage system run efficiently with low energy consumption as well as low costs. The success of this project will provide both theoretical methods and technical support for solving energy consumption problem of online social networks, and promote the progress of data storage technologies in large-scale online social networks.

面对在线社交网络广泛分布的用户群和巨大的数据量，大型社交网站纷纷将服务部署到全球多个数据中心。在倡导节能减排、绿色计算的今天，如此大规模社交网络存储系统的能耗问题开始引起学术界和工业界的广泛关注。现有研究一方面未充分结合在线社交网络特点，且主要针对访问性能和运行成本进行优化，较少考虑能耗因素；另一方面极少考虑诸如固态硬盘等高性能、低功耗存储介质的引入对能耗优化的积极作用。针对以上问题，本项目拟从物理资源与数据资源两个层面入手，在综合考虑能耗、成本和性能的基础上，结合在线社交网络在服务执行逻辑和交互方面的特点，对存储系统资源节能供给、多数据中心间数据节能布局和基于多存储介质的数据节能调度展开系统、深入的研究，以实现社交网络存储系统高效率、低能耗、低成本的运行。本项目的成功研究将对解决社交网络存储系统能耗问题提供理论方法和技术支撑，推动大规模在线社交网络数据存储技术的进步。

在线社交网络将现实生活中人们的社会关系与活动映射到网络虚拟空间，通过这种映射实现用户之间更加便捷高效地交流、信息共享和其他相关社会活动。伴随社交网络的蓬勃发展，其呈现出一些新的特点，如用户数据量与日俱增、访问呈现高并发性和用户间依据社会关系频繁交互等。这些特点对社交网络的持续发展提出巨大挑战，而应对这一挑战的基础是解决社交数据的存储问题，本项目正是以实现社交数据高效、低成本、绿色存储为核心目标。研究内容从两个层面展开：第一，从物理资源层面研究资源节能供给问题，为系统提供运行资源支持；第二，从数据资源层面研究数据在数据中心内和数据中心间节能存储问题。研究取得的成果简介如下：.首先，针对社交访问请求特点与执行逻辑，分析访问性能、资源使用成本和能耗之间关系，建立资源供给模型并设计优化算法，在保证延迟和负载均衡的情况下实现成本与能耗的权衡与优化，为大规模社交数据存储系统的构建提供资源保证。.其次，针对社交数据在数据中心内放置问题，从数据交互的访问特征、具体数据中心网络拓扑结构和存储介质等多方面，总结提出新的数据动态放置算法设计框架，从数据资源层面进一步为存储系统减少存储和网络开销，降低数据中心能耗。.再次，针对大型社交网站跨区域部署问题，在保证访问延迟的前提下，综合考虑成本与能耗，结合多种社交图设计数据节能放置策略，以减少存储、传输数据副本带来的成本和能耗的增加，再一次从数据层面为跨区域存储系统的高效、绿色运行提供技术保障。.项目组在ICDCS、JCST等国内外权威会议和期刊上发表和录用论文20篇，其中一篇论文获得大数据会议最佳论文奖，申请国家技术发明专利4项，获得软件著作权1项，协助培养了2名博士毕业生，4名硕士毕业生，圆满完成预期目标。