An Evaluation of the Google SPDY Protocol Vishal Patel Carnegie Mellon University vkpatel@andrew.cmu.edu ABSTRACT: The number of Internet users is growing constantly, leading to emerging web usage patterns that are increasingly large and complex. Researchers have been investigating various performance optimizations to improve the Hyper Text Transfer Protocol (HTTP), and Google's experimental SPeeDY (SPDY) protocol is a radical attempt at solving this problem by engineering the Session Layer of the TCP/IP protocol stack. It is designed to improve browsing experience by changing the way HTTP requests are made. The scope of this paper is to identify these enhancements, and to quantify their effects on the client side web traffic through assessment of latency and bandwidth measurements. Specifically, we aim to compare these enhancements with the existing HTTP protocol and verify if the benefits are compelling enough to promote SPDY as the successor of HTTP. From the results of our experiments, we conclude that although SPDY provides a reduction in latency of approximately 22% when compared to the existing HTTP protocol, it performs poorly in terms of network utilization. It is currently approximately 10% less efficient in utilizing the available packet size. Bearing in mind the magnitude of effort required in implementing a new session layer on the client and server side, SPDY needs to be more efficient in utilizing the contents of a packet, in order to justify this effort and emerge as the next standard for web communication.