Design of reliable and efficient communication and computing systems: architecture, code design, and optimization

三个重要的技术趋势奠定了现代信息经济的基础：无处不在的宽带数字通信的出现，我们从大量数据集中提取有用和可操作信息的能力的巨大进步，以及根据摩尔定律计算能力的持续扩展。在这些领域当前面临的挑战的推动下，我们将汇集这些领域的见解和最新进展，开展三个协同研究项目。第一个是架构理解，第二个是算法设计，第三个是更广泛的应用。第一个项目涉及如何最好地为流数据构建通信系统。无线行业预测，未来10年的容量需要增加1000倍。但是，尽管许多越来越重要的应用涉及实时内容交付，但这些数据的流性质与大多数通信系统的基于块的设计不匹配。我们将对这些系统的可靠性和延迟之间的权衡制定基本限制，并将开发所需的编码方法来实现这些限制。我们的第二个项目利用数学优化技术来开发解码纠错码的新算法。我们的方法有可能解决一些在纠错方面具有重要工业意义的长期问题。在我们的期末专题中，我们将研究大规模计算系统的功耗感知设计和操作。通过将排队知识和无线路由技术与电路和计算机体系结构的洞察力相结合，我们将开发用于大规模数据中心任务调度和电源管理的新算法。

Three important technological trends underlie the modern information economy: the emergence of ubiquitous broadband digital communications, great advances in our ability to extract useful and actionable information from enormous data sets, and the continued scaling of computational power according to Moore's Law. Motivated by current challenges in each of these areas, and drawing together insights and recent advances from across these areas, we will tackle three synergistic research projects. The first is in architectural understanding, the second in algorithmic design, the third in broader applications. The first project concerns how best to architect communication systems for streaming data. The wireless industry is forecasting the need for a 1000-fold increase in capacity over the next decade. But, while many applications of increasing importance involve real-time content delivery, the streaming nature of this data is ill-matched to the block-based designs of most communication systems. We will develop fundamental limits on the tradeoff between reliability and delay in these systems and will also develop the required coding approaches to achieve those limits. Our second project draws upon techniques of mathematical optimization to develop novel algorithms for decoding error-correcting codes. Our approach has the potential to resolve some long-standing issues of great industrial importance in error correction. In our final project we will investigate power-aware design and operation of large-scale computational systems. By combining knowledge of queuing and techniques of wireless routing with insight into circuits and computer architecture, we will develop new algorithms for task scheduling and power management in large-scale data centers.