Understanding the Impact of Compression on Web Retrieval Performance

JunLi Yuan, Institute for Infocomm Research, School of Computing (NUS), Singapore 119613. Email: junli@i2r.a-star.edu.sg

Xiang Li, School of Information Security, Shanghai JiaoTong University, Shanghai China 200030. Email: xiangli@sjtu.edu.cn

Chi-Hung Chi, School of Software, Tsinghua University, Beijing China 100084. Email: chich@comp.nus.edu.sg

Keywords

Web retrieval, latency, performance, compression, object latency, page latency.

Abstract

Compression is an important technique used to improve web retrieval latency. Many studies report the benefit of compression to be high. However, those studies mainly focus on object retrieval latency. In web access, page retrieval latency is more meaningful because the basic unit of web browsing is page instead of object. Thus it is important to study the impact of compression on page retrieval latency. In this paper, we investigate the factors that affect page latency and systematically study the impact of two different compression mechanisms on them. Our results show that pre-compression outperforms real-time compression in all situations, and the improvement on page latency achieved by compression is much lower than it does on object latency due to some factors specially found in page retrieval such as the dependency among objects in a page and the parallelism used in web retrieval.