Evolving the HTTP Ecosystem to Efficiently Support Video and other Large Files

点击翻译按钮获取中文摘要

Studies report that in North America video now accounts for 40-66% of all traffic on the Internet and it will rapidly increase to 90-95%. Internet video services are now using the same methods to deliver video to consumers as are used to deliver data from web searching, banking, on-line shopping, social networking and other popular services. The problem is that the existing systems used to deliver that content have not been designed and optimized to deliver video. The result is often a less than satisfying video viewing experience, with videos being interrupted or paused while the video is fetched from the server. This can be especially noticeable when trying to view high-definition videos or during times when others are using the same service or network. More importantly, there is evidence that shows that as the demand for video increases, it will make the existing services we have come to rely on slower and slower. Our research will devise new methods to improve the efficiency of the underlying systems software and networks used to deliver content to consumers. We will experimentally evaluate, identify bottlenecks in, and improve the performance of existing software systems. This will reduce the number of machines required by providers of Internet services, thereby reducing the power they consume and lowering their costs. For consumers this should reduce costs paid for these services. In addition, this will enable the smooth delivery of high-quality video to their desktop PC, laptop, netbook, tablet or smart phone whether they are at work, home, or on the road. Most importantly, we will make these improvements while ensuring that existing services that have become critical to our lives not only continue to function but function well.