CRII: NeTS: Optimizing Emerging Web Protocols for a Faster World Wide Web

点击翻译按钮获取中文摘要

This project seeks to optimize emerging web protocols for a faster world wide web (WWW). As the key protocol that supports WWW, the Hypertext Transfer Protocol (HTTP) has been very successful. It accounts for more than half of the overall Internet traffic. Currently, the most widely deployed HTTP version is HTTP/1.1 standardized 16 years ago. Nonetheless, several new web protocols ("NWP") such as HTTP/2, SPDY, and QUIC have been proposed recently to overcome limitations of HTTP/1.1. In particular, HTTP/2, the next version of HTTP, has been standardized in 2015 and started replacing HTTP/1.1. NWP will become prevalent in the foreseeable future. Given the importance of these web protocols, the objective of this project is to understand them, and to propose effective solutions for improving web performance by leveraging their common new features. Specifically, this project identifies key dimensions of NWP that were not explored before, and includes the following research tasks. (1) The principal investigator (PI) will conduct a comprehensive user study by collecting web traffic from voluntary participants, to understand NWP's performance "in the wild". The measurement results not only shed light on improving the performance, energy efficiency, and mobile-friendliness of current NWP, but also provide benchmarks for evaluating future NWP. (2) The PI will investigate how to strategically leverage Server Push, a new feature of many NWP, to improve web performance. Server push allows a server to preemptively send (push) resources to the client, thus overcoming many limitations of the conventional "request-response" resource fetching paradigm. (3) The PI will study how to optimize NWP performance by simultaneously leveraging multiple interfaces (e.g., Wi-Fi and cellular on mobile devices). A new framework promises to allow multipath to be adaptively utilized only when necessary, in order to reduce the energy consumption and cellular data usage for mobile devices. (4) The PI also explores mechanisms for improving the interaction between NWP and the underlying transport layer such as the Transmission Control Protocol (TCP).Broader Impacts: Successful completion of this research will benefit the society, due to the importance of web protocols in our everyday computing. It is known that Internet users are very sensitive to web page load time. The proposed solutions will improve web users' experience, and they can be deployed on browser and server applications to yield immediate benefits. Note the proposed solutions do not rely on a specific web protocol, and can therefore yield long-term benefits. Dissemination of the research results will be done through academic publications, code/data distribution, industrial collaborations, and standardization. The PI has included HTTP/2 in the course he taught, and will incorporate more NWP materials into the undergraduate and graduate curriculum.