A Design Framework for Improving Reliability, Debug and Security of Multi-Core Systems

“该奖项是根据2009年美国复苏和再投资法案（公法111-5）资助的。“项目编号：0903191 PI姓名：Sandip Kundu和Israel Koren标题：一个提高多核系统可靠性、可扩展性和安全性的设计框架Inst：马萨诸塞州大学，阿默斯特摘要：今天？的计算机芯片的特点是在同一个衬底上有多个内核，所有内核都可以执行并发任务，从而提高了整体性能。然而，多核系统加剧了一些问题，如功耗，系统和软件调试，安全性和长期系统可靠性。计算机系统中的功耗是主要关注的问题。例如，谷歌服务器工厂，据报道消耗约103兆瓦的电力？足以为华盛顿的塔科马大小的城市供电。多任务、多线程、多核环境中的多核硬件/软件非常复杂，因为单个核可能崩溃并且通信可能死锁。随着社会越来越依赖计算机系统的可用性以及它们处理和存储的数据的安全性，计算机的可靠性和安全性越来越受到关注。因此，持续监察电脑系统的健康状况，以便在故障发生后迅速复原，并保护重要及私人资料免受窜改，是十分必要的。该计划的构思是设计一个简单的专用监控核心，并将其集成于晶片多处理器内，使其永远运作。该核心支持执行用于重新启动各个核心的命令，并允许查询同一基板上的任何核心的内部状态。这样的核心几乎不增加系统成本，但增加了许多运行时功能，使我们能够解决上述问题，并允许改进远程系统管理，这已经成为计算机数据中心的一个主要挑战。在技术层面，它提供解决方案，以减轻社会问题，如能源消耗，提高计算机安全性和改善计算机正常运行时间。在课程层面，它培养了一门对美国工业竞争力极其重要的学科的研究人员。该项目通过技术合作得到了美国主要行业的积极支持。

"This award is funded under the American Recovery and Reinvestment Act of 2009(Public Law 111-5)."Project ID: 0903191PI names: Sandip Kundu and Israel KorenTitle: A Design Framework for Improving Reliability, Debug and Security of Multi-Core SystemsInst: University of Massachusetts, AmherstABSTRACT:Today?s computer chips feature multiple cores in the same substrate where all cores may perform concurrent tasks, improving overall performance. However, multi-core systems exacerbate some problems such as power dissipation, system and software debugging, security and long term system reliability. Power consumption in computer systems is a major concern. For example, Google server plants, reportedly consume about 103 megawatts of electricity ? enough to power a city of the size of Tacoma, Washington.Debugging multi-core hardware/software in a multitasking, multithreading, multi-core environment is very complex as an individual core may crash and communication may deadlock. Computer reliability and security are increasing concerns, as society becomes ever more dependent on the availability of computer systems and on the security of the data that they process and store. Continuous monitoring of the health of the computer system to allow fast recovery from faults, and protecting critical and private data from tampering attempts are thus necessary.The project envisions designing a simple dedicated supervisory core integrated into chip multi-processors that is always alive. This core supports execution of commands for rebooting individual cores and allows query of internal states of any core on the same substrate. Such a core adds little to system cost but adds a number of run time capabilities that allow us to solve the above problems and also allow improved remote system management, which has emerged as a major challenge in computer data centers.The broader impact of this project is two fold. At a technical level, it provides solutions to mitigate societal problems such as energy consumption, advancing computer security and improving computer up-time. At a curriculum level, it trains researchers in a discipline that is extremely important for US industrial competitiveness. This project is actively supported by major US industries through technical collaboration.