1. 1 Building Reliable Web Services: Methodology, Composition, Modeling and Experiment Pat. P. W. Chan Department of Computer Science and Engineering The Chinese University of Hong Kong 23 October 2007

2. 2 Outline  Introduction to Web Services  Problem Statement  Methodologies for Web Service Reliability  New Reliable Web Service Paradigm  Web Service Composition Algorithm  Experimental Results and Discussion  Conclusion

3. 3 Introduction  Service-oriented computing is becoming a reality.  The problems of service dependability, security and timeliness are becoming critical.  We propose experimental settings and offer a roadmap to dependable Web services.

4. 4 Problem Statement  Fault-tolerant techniques Replication Diversity  Replication is one of the efficient ways for providing reliable systems by time or space redundancy. Increasing the availability of distributed systems Key components are re-executed or replicated Protect against hardware malfunctions or transient system faults  Another efficient technique is design diversity Employ independently designed software systems or services with different programming teams, Defend against permanent software design faults.  We focus on the analysis of the replication techniques when applied to Web services.  A generic Web service system with spatial as well as temporal replication is proposed and investigated.

5. 5 Proposed Paradigm Replication Manager RR Algorithm / Voting WatchDog UDDI Registry WSDL Web Service IIS Application Database Web Service IIS Application Database Web Service IIS Application Database Client Port Application Database Register Look up Get WSDL Invoke web service Keep check the availability of all the web. If Web service failed, update the list of availability of Web services Update the WSDL

6. 6 Round Robin

7. 7 Parallel N-Version Programming

8. 8 Recovery Block

9. 9 Experiments  A series of experiments are designed and performed for evaluating the reliability of the Web service, Spatial replication00001111 Reboot00110011 Retry01010101

10. 10 Testing system  Best Route Finding.  Provide traveling suggestions for users.  Starting point and destination.  The system needs to provide the best route and the price for the users.

11. 11 System Architecture

12. 12 Experimental Setup  Examine the computation to communication ratio  Examine the request frequency to limit the load of the server to 75%  Fix the following parameters Computation to communication ratio (e.g 10:1) Request frequency

13. 13 Experimental Setup Communication time: Computation time 143:14 (10:1) Request frequency1 request per min Load78.5% Timeout period of retry1 min Timeout for Web service in RM1s (web service specific) Polling frequency10 requests per min Number of replicas5 Max number of retries5 Round-robin rate1 s

14. 14 Experiment Parameters  Fault mode Temporary (fault probability: 0.01) Permanent (fault probability: 0.001)  Experiment time 5 days (7200 requests)  Measure: Number of failures Average response time (ms)  Failure definition: 5 retries are allowed. If there is still no correct result from the Web service after 5 retries, it is considered as a failure.

15. 15 Experimental Result with Round-robin (failures / response time in ms) ExperimentsSingle serverSingle server with retry Single server with reboot (continues no response for 3 requests) Single server with retry and reboot Spatial Replication RR Hybrid approach RR+Retry Hybrid approach RR+ Reboot All round approach RR spatial + Retry (5 times) + reboots Normal case0 / 1830 / 1930 / 1900 / 1870 / 1880 / 1950 / 1930 / 190 Temp705 / 1900 / 223723 / 2310 / 238711 / 1870 / 233726 / 1880 / 231 Perm6144 / --6337 / --1064 / --5 / 25785637 / --5532 / --152 / 1870 / 191

16. 16 Experimental Result with N-Version (failures / response time in ms) ExperimentsSingle serverSingle server with retry Single server with reboot (continues no response for 3 requests) Single server with retry and reboot Spatial Replication Voting Hybrid approach Voting+ Retry Hybrid approach Voting + Reboot All round approach Voting spatial + Retry (5 times) + reboots Normal case0 / 1830 / 1930 / 1900 / 1870 / 1890 / 1900 / 188 Temp705 / 1900 / 223723 / 2310 / 2380 / 190 0 / 1890 / 187 Perm6144 / --6337 / --1064 / --5 / 25783125 / 1913418 / 19240 / 1890 / 188

17. 17 Experimental Result with Recovery Block (failures / response time in ms) ExperimentsSingle serverSingle server with retry Single server with reboot (continues no response for 3 requests) Single server with retry and reboot Spatial Replication Voting Hybrid approach Voting+ Retry Hybrid approach rollback + Reboot All round approach rollback spatial + Retry (5 times) + reboots Normal case0 / 1830 / 1930 / 1900 / 1870 / 1910 / 1890 / 1930 / 188 Temp705 / 1900 / 223723 / 2310 / 2380 / 2050 / 2030 / 2040 / 201 Perm6144 / --6337 / --1064 / --5 / 25783478 / 2153245 / 208201 / 2110 / 201

18. 18 Web Service Composition Algorithm  N-version programming Reliable Efficient  Composition WSDL WSCI  Verification BPEL Petri-Net

19. 19 … … WSDL

20. 20 <action name="ReceiveStartpointDest “ role="tns:busAgent “ operation="tns:BRF/shortestpath"> <action name="Receivecheckpoint “ role=" tns:busAgent “ operation="tns:BRF/addCheckpoint"> … WSCI

21. 21

22. 22 1.Output 2.Operation in WSDL 3.Find the output information in CP1 (Web service component) 4.If Input of the operation == required input CP1 5.Else search in the WSCI of CP1 to find action == operation 6.Get the pervious action involved 7.Search in WSDL to find operation == action 8.If Input of the operation == required input CP4 Else, till the root of WSCI Web service composition CP7 CP10

23. 23

24. 24

25. 25 Petri-Net – Basic Activities

26. 26 Petri-Net – Structure Activities

27. 27 Composed Petri-Net

28. 28

29. 29

30. 30

31. 31

32. 32

33. 33

34. 34 Petri-Net (Four identical replicas)

35. 35 Petri-Net (N-version Web service with voting)

36. 36 Petri-Net (Recovery Block)

37. 37 (a) (b) P1P1 λ1λ1 μ1c2μ1c2 S-j P2P2 μ2c2μ2c2 λ2λ2 S-j-1 S S-n F λNλN μ*c 2 (1-c 1 )μ* λ*λ* S-1S-2 λ*λ* μ*c2μ*c2 λ*λ* (1-c 1 )μ* F (1-c 1 )μ 1 (1-c 1 )μ 2 (1-c 2 )μ 1 (1-c 2 )μ 2 Reliability Model

38. 38 Reliability Model IDDescriptionValue λnλn Network failure rate0.02 λ*Web service failure rate0.025 λ1λ1 Resource problem rate0.142 λ2λ2 Entry point failure rate0.150 μ*Web service repair rate0.286 μ1μ1 Resource problem repair rate0.979 μ2μ2 Entry point failure repair rate0.979 C1C1 Probability that the RM response on time0.9 C2C2 Probability that the server reboot successfully0.9

39. 39 Outcome (SHARPE)

40. 40 Conclusion  Surveyed replication and design diversity techniques for reliable services.  Proposed a hybrid approach to improving the availability of Web services.  Proposed a Web service composition algorithm and verified by Petri-Net.  Carried out a series of experiments to evaluate the availability and reliability of the proposed Web service system.

41. 41 Q&A