CSR: Designing Next Generation Data-Centers with Advanced Communication Protocols and Systems Services

该项目将为分布式计算机和数据密集型应用环境开发技术，如电子商务、娱乐、金融、搜索引擎、医疗信息学、基因组学和基于网络的内容交付中遇到的技术。这些新兴应用中的许多本质上都是数据密集型的。数据是通过交易、工具和模拟等多种来源产生的。不仅数据集的大小在增加，而且它们正变得更加动态和分布式。当前数据中心使用的设计有几个局限性，无法以较低的成本实现良好的性能和可扩展性，以处理动态和地理分布数据的新兴趋势。该项目将围绕计算机体系结构、网络、集群计算和数据等方面展开研究。该项目将沿着以下方向开展研究：利用现代网络的新机制和新特性设计先进的通信协议和子系统；设计和开发新的数据中心原语，如软共享状态、分布式锁管理器和全局内存聚合器；设计动态内容缓存服务；以及设计基于约束的主动资源适配服务，以提供服务质量保证。按照建议的方向开发的解决方案计划集成在实验数据中心测试床和模拟测试床上，并使用一组商业工作负载进行评估。

The project will develop technologies for distributed computer and data intensive application environments such as those encountered in e-commerce, entertainment, financial, search engines, medical informatics, genomics, and web-based content delivery. Many of these emerging applications are quite data-intensive in nature. Data is being generated from multiple sources such as transactions, instruments, and through simulations. Not only datasets are increasing in size, they are becoming more dynamic and distributed. The designs used by current-day data-centers have several limitations to achieve good performance and scalability with low cost for handling the emerging trend of dynamic and geographically distributed data. The proposed research is built around computer architecture, networking, cluster computing, and data.The project will carry out research along the following directions: designing advanced communication protocols and subsystems by taking advantage of novel mechanisms and features of modern networks; designing and developing novel data-center primitives such as soft shared state, distributed lock managers and global memory aggregators; designing dynamic content caching services; and designing constraint-based active resource adaptation services to provide Quality of Service (QoS) guarantees. Solutions developed along the proposed directions are planned to be integrated on an experimental data-center testbed and a simulated testbed and evaluated using a set of commercial workload.