|
QW2002 Paper 2I1 Dr. Ludmila Cherkasova, Dr. Yun Fu, Dr. Wenting Tang & Dr. Amin Vahdat Web Site Performance and Qos Monitoring |
|
Understanding and measuring end-to-end service performance perceived by the clients is a challenging task. In this work, we describe a novel technique for measuring the web site's end-to-end response time which based on reconstruction of web page accesses from passively captured network packets at a web server side. This technique can effectively determine the set of objects composing web page without parsing or interpreting the HTML syntax. Exploiting this technique, we built a tool, called EtE monitor. EtE monitor does not require any changes or modifications to site content or server side infrastructure, or client browsers, and can be used for sites with static or dynamically generated content.Relative to existing approaches, EtE monitor offers a set of new benefits. Timestamps extracted from network packet level provide invaluable information about connection setup time, server side processing time, and network related transfer time, which can not be directly obtained from other sources. Additionally, this technique allows us to analyze many other useful and practical metrics, such as number of aborted page accesses, number of objects retrieved directly from the server versus from network and client browser caches, number of client connections used to retrieve the web page, amount of packets resent in the response to reflect network congestion on the path to a client, etc.
Ludmila Cherkasova joined Hewlett-Packard Laboratories, Palo-Alto, California, in 1991, where she currently is a senior scientist in the Internet and Computing Platforms Research Center. Among her latest works are WebMetrix (a tool for web service providers), MediaMetrics (a tool for media service providers), content-aware load balancing strategies for web server clusters, a scaleable architecture for shared web hosting service, the admission control and scheduling algorithms that improve performance of overloaded web servers. Her research interests include the performance of Internet-based services, computer systems, network protocols, and characterization of next generation system workloads. She received a PhD in Computer Science from Novosibirsk Institute of Computing Systems, Russia, in 1984. She is a member of ACM and USENIX.Yun Fu is a Ph.D candidate at Computer Science Department of Duke University. He received his B.S. in computer science and Master degree in computer engineering from Nankai Univeristy in 1995 and 1998 respectively, and his MS in computer science from Duke University in 2001. His research interests include distributed systems, networking and operating systems.
Wenting Tang is a member of technical staff in Internet System and Storage Laboratory, Hewlett Packard, Palo Alto, California. His research interests include Internet service performance, content delivery networks, streaming media and large-scale distributed systems. He received his PhD in computer science from Michigan State University. He is a member of the IEEE.
Amin Vahdat joined the faculty of the the Computer Science department at Duke University after receiving a PhD from the University of California, Berkeley in 1998. His research focuses on system support for highly available and high performance network services. As a graduate student he participated in the design and implementation of the UC Berkeley Network of Workstations project. Some of this work was later commercialized as part of the Inktomi search engine. At Duke University, A. Vahdat leads the Internet Systems Software Group (ISSG). A. Vahdat is a winner of the prestigious National Science Foundation Faculty Development CAREER award and a member of ACM, IEEE, and USENIX.