1. 1 PREFETCHING INLINES TO IMPROVE WEB SERVER LATENCY Ronald Dodge US Army rdodge@gmu.edu Daniel Menascé, Ph. D. George Mason University menasce@cs.gmu.edu  Dodge & Menascé. All Rights Reserved.

2. 2 Electronic Commerce: online sales are soaring “… IT and electronic commerce can be expected to drive economic growth for many years to come.” The Emerging Digital Economy, US Dept. of Commerce, 1998.  Dodge & Menascé. All Rights Reserved.

3. 3 Business in the Internet Age (Business Week, June 22, 1998)

4. 4 Caution Signs Along the Road There will be jolts and delays along the way for electronic commerce: congestion is the most obvious challenge. ( Gross & Sager, Business Week, June 22, 1998, p. 166.)

5. 5 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

6. 6 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

7. 7 Response Time Reduction Techniques Caching –web browser –proxy server Prefetching –predict what documents will be requested and bring them into the cache ahead of time.  Dodge & Menascé. All Rights Reserved.

8. 8 Browser Caching  Dodge & Menascé. All Rights Reserved.

9. 9 Browser Caching 1 2 3 4  Dodge & Menascé. All Rights Reserved.

10. 10 Browser Caching 3 4 1 2  Dodge & Menascé. All Rights Reserved.

11. 11 Browser Caching 3 4 1 2  Dodge & Menascé. All Rights Reserved.

12. 12 Caching Proxy Server 1 2 3 4  Dodge & Menascé. All Rights Reserved.

13. 13 Caching Proxy Server 1 2 3 4  Dodge & Menascé. All Rights Reserved.

14. 14 Caching Proxy Server 1 2 3 4  Dodge & Menascé. All Rights Reserved.

15. 15 Caching Proxy Server 1 2 3 4  Dodge & Menascé. All Rights Reserved.

16. 16 Caching Proxy Server 1 2 3 4  Dodge & Menascé. All Rights Reserved.

17. 17 Caching Proxy Server 1 2 3 4  Dodge & Menascé. All Rights Reserved.

18. 18 Goal Improve user perceived latency through caching and prefetching of inlines at the server side.  Dodge & Menascé. All Rights Reserved.

19. 19 Related Work on Prefetching Padmanabhan, 1995 –effects of prefetching on latency and network load. Chinen and Yamaguchi, 1996 –prefetching to the proxy server the first N links of any Web page. Crovella and Badford, 1997 –network load can be minimized by adjusting prefetch rate Foxwell and Menasce, 1998 –prefetching results of search engine queries.  Dodge & Menascé. All Rights Reserved.

20. 20 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

21. 21 Browser Server HTTP request inline 1 request inline 2 request HTTP document inline 1 file inline 2 file HTML document parsed by the browser server disk No Caching/Prefetching of Inlines  Dodge & Menascé. All Rights Reserved. www.cs.gmu.edu

22. 22 Capacity Planning for Web Performance The introduction: [ Introduction ] [ Capacity Planning ] [ Web Performance Problems ] [ Web Performance Modeling ] [ Publications ]  Dodge & Menascé. All Rights Reserved. inline

23. 23 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

24. 24 Caching/Prefetching of Inlines  Dodge & Menascé. All Rights Reserved. www.cs.gmu.edu

25. 25 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

26. 26 Workload Characterization Analysis of the HTTP log of a Web server at GMU: - requester’s address - request arrival time - type of request - file requested - size of file requested  Dodge & Menascé. All Rights Reserved.

27. 27 Previous Workload Characterization Work Arlitt and Williamson (1996), Crovella and Bestravos (1996) The average size of a transferred document does not exceed 21KB Less than 3% of the requests are for distinct files. The file size distribution is heavy-tailed. File inter-reference times are exponentially distributed and independent.  Dodge & Menascé. All Rights Reserved.

28. 28 Interrarival time distribution  Dodge & Menascé. All Rights Reserved. average = 2 sec.

29. 29 HTML file size distribution  Dodge & Menascé. All Rights Reserved. average = 3,552 bytes.

30. 30 Distribution of number of inlines per HTML document  Dodge & Menascé. All Rights Reserved. average = 2.92 inlines.

31. 31 Inline file size distribution  Dodge & Menascé. All Rights Reserved. average = 10,392 bytes.

32. 32 File Interreference A.html B.html C.html D.html A.html interreference distance = 3 Repeat file? Use interreference distribution to decide how many requests to look back. YES Pick a new file NO  Dodge & Menascé. All Rights Reserved.

33. 33 File interreference distribution  Dodge & Menascé. All Rights Reserved.

34. 34 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

35. 35 Network CPU Web browsers WEB Server Cache Disk h 1 - h Simulation Environment:  Dodge & Menascé. All Rights Reserved.

36. 36 Simulation Experiment Simulation Tool: CSim Batch means analysis: accuracy 10% and confidence level of 95%. Cases studied:  Dodge & Menascé. All Rights Reserved.

37. 37 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

38. 38 Response time (in sec) of HTML document requests vs. cache size (in Kbytes)  Dodge & Menascé. All Rights Reserved.

39. 39 Response Time of Inline Files (in sec) vs. Cache Size (KB)  Dodge & Menascé. All Rights Reserved.

40. 40 Total Response time (in sec) vs. cache size (in Kbytes)  Dodge & Menascé. All Rights Reserved.

41. 41 File Hit Ratio vs. cache size (in Kbytes)  Dodge & Menascé. All Rights Reserved.

42. 42 Byte Hit Ratio vs. cache size (in Kbytes).  Dodge & Menascé. All Rights Reserved.

43. 43 Server disk utilization vs. cache size (in Kbytes).  Dodge & Menascé. All Rights Reserved.

44. 44 Total Response Time (sec) vs Cache Size (KB) for Prefetching and non-Prefetching Cases  Dodge & Menascé. All Rights Reserved. caching used in both cases

45. 45 Outline Response Time Reduction Techniques HTTP Transactions without Prefetching HTTP Transactions with Prefetching Model –Workload characterization –Simulation Model Results Concluding Remarks  Dodge & Menascé. All Rights Reserved.

46. 46 Concluding Remarks Prefetching can be used in conjunction with caching to decrease user perceived latency. No modification to the HTTP protocol nor the browser software is needed. Results showed a 48% improvement in response time through prefetching of inlines.  Dodge & Menascé. All Rights Reserved.

47. 47 Relevant Bibliography Menascé, D. A. and V. A. F. Almeida, Capacity Planning for Web Performance: metrics, models, and methods,” Prentice Hall, Upper Saddle River, NJ, 1998. N. Yeager and R. McGrath, Web Server Technology, Morgan Kaufmann, San Francisco. 1996. Foxwell, H. and D. A. Menascé, Prefetching Results of Web Searches, Proc. of the 1998 Computer Measurement Group Conference, Anaheim, CA, Dec. 6-11, 1998.  Dodge & Menascé. All Rights Reserved.