Browser workload characterization for an Ajax-based commercial online service

  • Authors:
  • Shu Xu; Bo Huang; Junyong Ding; Jinquan Dai

  • Affiliations:
  • Intel China Software Center, China;Intel China Software Center, China;Intel China Software Center, China;Intel China Software Center, China

  • Venue:
  • IISWC '09 Proceedings of the 2009 IEEE International Symposium on Workload Characterization (IISWC)
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

The transition to cloud computing and SaaS is a disruptive trend where users can conveniently access the services through browsers at any clients. In addition, with the prevalence of Web 2.0 and AJAX techniques, a browser-based client can have complex application logic and fancy user interface that are comparable to traditional desktop applications. This paper reports the study of workload construction and characterization for browser-based clients, using the Ajax-based web client of Zimbra (a commercial online messaging and collaboration suite). By comparing the various workload behaviors across different Zimbra server datasets, different browsers and different client platforms, it presents the characteristics of a real-life web application, which has significant differences from existing browser benchmarks in the literature. In addition, the platform-independent and browser-independent design of our workload makes it portable across various clients. Finally, this paper also provides valuable insights to the browser internals by analyzing the workload execution, the browser memory footprint and the breakdown of browser sub-modules.