Low-Latency Streaming and Storage Systems using Error Correcting Codes

未来的互联网预计将支持各种低延迟沉浸式多媒体应用，包括高端远程呈现、增强/虚拟现实和在线游戏。不幸的是，最先进的通信和存储系统仍然没有达到这些应用程序中严格的延迟目标。拟议的研究将开发新的纠错码(ECC)，用于支持通信和存储系统，这些系统可以实现多媒体应用程序延迟的数量级改进。虽然ECC在通信和存储系统中都有丰富的历史，但无论是在学术研究还是现实系统中，它们在低延迟流应用中的应用都没有引起太大的兴趣。相反，传统的ECC设计涉及非常长的块长度，这可能会引入显著的延迟。这种ECC在低延迟应用中的幼稚应用将极大地损害它们的纠错能力，或者显著增加延迟。我们认为这样的ECC不应该用于流媒体应用程序。然而，如果从基本原理明智地开发新的编码方案，则可以实现显著的改进。为此，我们将使用信息论中的技术来研究ECC在流应用中的基本限制。尽管几十年来，这样的研究一直是一个悬而未决的问题，但最近，PI和他的合作者通过在这个问题上发展了一条新的攻击线，取得了重大突破。使用这种方法，拟议的研究将在流应用的低延迟ECC领域做出几项开创性的贡献。此外，拟议的研究将是极少数利用机器学习方法来设计不断适应不断变化的网络条件的ECC的努力之一。当人们考虑点播流媒体应用时，通过在网络边缘缓存流行内容来减少延迟的主要机会存在。新兴的无线网络将由基站和接入点组成的超高密度网络组成，并为它们提供在本地存储数据的能力。通过在这些系统中应用ECC，我们可以将单个设备上的存储转变为单一的虚拟缓存，具有显著更高的存储效率和可靠性。拟议的研究将利用存储系统(包括分布式存储代码和编码缓存)中ECC的最新进展，从基本原理出发在存储系统中开发延迟优化的ECC，并将其与提案第一部分讨论的流通信系统的ECC集成。建议的研究有可能大幅降低直播和点播流应用的成本，通过以较低成本支持高端网真通信来减少商务旅行，并为加拿大提供环境和经济利益。

The future Internet is expected to support a variety of low-latency immersive multimedia applications including high-end telepresence, augmented/virtual reality and online gaming.  Unfortunately state-of-the-art communication and storage systems still fall short of meeting the stringent latency targets in these applications. The proposed research will develop new  error correcting codes (ECC)  for supporting communication and storage systems that can achieve orders of magnitude improvement in latency for multimedia applications. Although ECC have a rich history in both communication and storage systems, their utility in low-latency streaming applications has not received much interest in either academic research or real world systems. On the contrary, traditional ECC designs involve very long block-lengths which can introduce significant delays. A naive application of such ECC in low-latency applications will drastically compromise their error correction capability, or significantly increase latency. We argue that such ECC should not be used in streaming applications. However if one judiciously develops new coding schemes, from first principles, then significant improvements can be achieved. Towards this end, we will study fundamental limits of ECC in streaming applications using techniques from information theory. Although such a study has remained an open problem for several decades, recently the PI and his collaborators have achieved  major breakthroughs by developing a new line of attack on this problem. Using this methodology, the proposed research is poised to make several pioneering contributions in the area of low-latency ECC for streaming applications. In addition, the proposed research will be one of the very few efforts to make use of machine learning methods in designing ECC that continuously adapt to changing network conditions. When one considers on-demand streaming applications, a major opportunity for reducing latency exists by caching popular content at the network edge. Emerging wireless networks will consist of an ultra dense network of base stations and access points and provide them with the ability to store data locally. By applying ECC in these systems, we can turn the storage at individual devices into a  single virtual cache with significantly better storage efficiency and reliability. The proposed research will leverage on recent advancements in ECC in storage systems including distributed storage codes and coded caching,  develop latency-optimized ECC in storage systems from first principles, and integrate them with ECC for streaming communication systems discussed in the first part of the proposal.  The proposed research has the potential to dramatically reduce cost of both live streaming and on-demand streaming applications, reduce business travel by supporting high-end telepresence communication at a lower cost, and provide both environmental and economic benefits to Canada.