Service differentiation is a way for website operators to provide better throughput and shorter user-perceived latency to some specific users. This work presents a request scheduling algorithm deployed at the website gateway to enable the Web quality of service without modifying client’s or server’s software. The QoS website gateway consists of a request classifier, a request scheduler, and a server prober. Unlike traditional packet scheduling algorithms which are mostly work-conservative, our algorithm appears to be non-work-conservative for access link; but is work-conservative for the server and the reverse direction because the service time of a request depends on the size of its response, not the size of the request itself. We emulate DRR and the window control mechanism to decide the order and release time of requests, respectively. The order is according to the response size of the requests and the pre-defined service weights. The release time is according to the service rate of the Web server. The HTTP requests incoming to the gateway will be classified and accumulated into the corresponding queues by the content-aware request classifier according to the pre-defined QoS policies. The deficit round robin scheduler with the window control mechanism decides which request should be fetched next and when it should be released to the Web server. The server prober scans URL, gets the corresponding response size of the Web pages on the server, and feeds the probed results to the request scheduler for helping the scheduling.
The QoS website gateway is evaluated with fixed-size Web pages and mixed-size Web pages to demonstrate the robustness of our approach. The throughput and the user-perceived latency of each service class are measured to demonstrate the effect of the service differentiation. The evaluation is performed with three service classes whose weight ratio is set to 6:3:1, and various Web pages whose sizes are ranging from 32 bytes to 128K bytes.
Under the QoS-disabled case, the three service classes get almost the same throughputs and the user-perceived latencies (1:1:1). On the contrary, under the QoS-enabled case, the three service classes get expected throughputs and user-perceived latencies as the pre-defined service weights (6:3:1), regardless whatever page sizes. In our evaluation, the throughput of the class with the largest weight is improved by up to 176%, while that of the class with the smallest weight is penalized by 52%. On the other hand, the user-perceived latency of the former is improved by up to 69%, while that of the latter is penalized by 75%.
There are mainly two directions for future works. The Web pages on a Web server can be static or dynamically generated. The URL and the response size of a static page are easy to seize and are used for the scheduling. A dynamically generated Web page varies its URL and response size. When dealing with dynamical Web pages, the QoS website gateway has to be equipped with a more sophisticated server prober to correctly estimate the response size.
Furthermore, enabling service differentiation at the QoS website gateway for a cluster of servers is also a considerable work. In this case, the QoS website gateway has to schedule the requests and balance the server load simultaneously.
R EFERENCES
[1] Keynote Systems Inc., http://www.keynote.com
[2] P. Mills and C. Loosley, “A performance Analysis of 40 e-Business Web Sites,” CMG J.
of Computer Resource Management, issue 102, 2001.
[3] R. Braden, D. Clark, and S. Shenker, “Integrated Services in the Internet Architecture: an Overview,” RFC 1633, June 1994.
[4] S. Blake, D. Black, M. Carlson, E. Davies, Z. Wang, and W. Weiss, “An Architecture for Differentiated Service,” RFC 2475, Dec. 1998.
[5] R. Braden, Ed., L. Zhang, S. Berson, S. Herzog, and S. Jamin, “Resource ReSerVation Protocol (RSVP) -- Version 1 Functional Specification,” RFC 2205, Sep. 1997.
[6] J. Almeida, M. Dabu, A. Manikutty, and P. Cao, “Providing Differentiated Levels of Service in Web Content Hosting,” in Workshop Internet Server Performance, Madison, WI, June 1998.
[7] R. Pandey, J. F. Barnes, and R. Olsson, “Supporting Quality of Service in HTTP Servers,” in Proc. 7th Annu. ACM Symp. Principles of Distributed Computing, Puerto Vallarta, Mexico, June 1998, pp. 247-256.
[8] L. Eggert and J. Heidemann, “Application-Level Differentiated Services for Web Servers,” World Wide Web J., vol. 2, no. 3, pp. 133-142, Aug. 1999.
[9] N. Bhatti and R. Friedrich, “Web Server Support for Tiered Services,” IEEE Network, vol. 13, issue 5, pp. 64-71, Sep.-Oct. 1999.
[10] N. Vasiliou and H. Lutfiyya, “Providing a Differentiated Quality of Service in a World Wide Web Server,” ACM SIGMETRICS Performance Evaluation Review, vol. 28, issue 2, pp. 22-28, Sep. 2000.
[11] X. Chen and P. Mohapatra, “Performance Evaluation of Service Differentiating Internet
Servers,” IEEE Trans. Computers, vol. 51, issue 11, pp. 1368-1375, Nov. 2002.
[12] M. Shreedhar and G. Varghese, “Efficient Fair Queuing Using Deficit Round-Robin,”
IEEE/ACM Trans. Networking, vol. 4, issue 3, pp. 375-385, June 1996.
[13] C. Semeria, “Supporting Differentiated Service Classes: Queue Scheduling Disciplines,”
Juniper Networks Inc., Sunnyvale, CA, Part Number: 200020-001, Dec. 2001.
[14] A. Demers, S. Keshav, and S. Shenker, “Analysis and Simulation of a Fair Queueing Algorithm,” in Proc. ACM SIGCOMM Computer Communication Review, vol. 19, issue 4, Austin, TX, Aug. 1989, pp. 1-12.
[15] M. Katevenis, S. Sidiropoulos, and C. Courcoubetis, “Weighted Round-Robin Cell Multiplexing in a General-Purpose ATM Switch Chip,” IEEE J. Selected Areas in Communications, vol. 9, issue 8, pp. 1265-1279, Oct. 1991.
[16] M. Murata, S. St. Laurent, and D. Kohn, “XML Media Types,” RFC 3023, Jan. 2001.
[17] The NetBSD Project, http://www.netbsd.org
[18] IP Filter - TCP-IP Firewall-NAT Software, http://coombs.anu.edu.au/~avalon [19] The Apache HTTP Server Project, http://httpd.apache.org
[20] WebBench, http://www.veritest.com/benchmarks/webbench
[21] P. Barford and M. Crovella, “Generating Representative Web Workloads for Network and Server Performance Evaluation,” ACM SIGMETRICS Performance Evaluation Review, vol. 26, issue 1, pp. 151-160, June 1998.