1. Architecture-based Reliability of web services Presented in SRG Group meeting January 24, 2011 Cobra Rahmani

2. Agenda Definition Architecture-based reliability modeling Research problem Challenges Our Approach

3. (Architecture-based reliability modeling) "Architecture-based reliability modeling estimates system reliability taking into account the information about the architecture of the software made out of components" [1] [1] K. Goseva–Popstojanova, A. P. Mathur, and K. S. Trivedi, “Comparison of Architecture-Based Software Reliability Models”, 12th Int’l Symp. on Software Reliability Engineering, 2001, pp. 22-31. Prevalent architecture-based analysis techniques Path-based techniquesState-based techniques Definition

4. The primary objective of architecture-based software reliability analysis: Component reliabilitiesApplication architecture to obtain an estimate of the application reliability based on Definition

5. Web service reliability modeling using SPN Service Provider Web Server Application Server Web Service Data Base SOAP/HTTP Request Shared Resources SOAP/HTTP Response Resource releaseResource allocation Middleware Http thread pool EJB pool DB connection pool Research problem

6. Web services is not standalone software and deployed into the servers Need to define Components from web services as well as web server and application server Challenges

7. Challenges Since web services run on middleware, some failures in service-based software are originating from interconnection of layers and not from any specific component or layer. (e.g. misconfiguration of shared resources)

8. Our Approach Start Static analysis 3. Extract main layers and shared resources 4. Build interaction between layers Instrumentation 1. Instrument AS with a profiler 2. Execute WS for each test case Dynamic Analysis 5. Estimate Transition Probabilities 6. Estimate Time Spent in each layer 7. Gather failure data for each layer 8. Form Petri net model 9. Run Petri net model, gather and analyze results End

9. Instrumentation Web server Web service Application server

10. Static Analysis Major layers and shared resources between layers will be extracted Layers: Web server, Application server, Web service, DB Shared Resources: Http/AJP thread pool in web container, EJB instance pool in EJB container, and Data Base connection pool in data access layer

11. Estimate Transition Probabilities ◦ Using JRAT, … ◦ Limitation: Output can be read from JRAT framework, manual work Estimate Time Spent in each layer ◦ Using JRAT, … Gather failure data for each layer ◦ Test cases, Bug repositories Dynamic Analysis

12. Form Petri net model

13. Gather reliability data from Petri net Sensitivity analysis Evaluation of this model with actual reliability data (tools: Jmeter, …) Form Petri net model

14. Further work Experimental analysis to find out what are the major failure types of web services (based on some data released on Internet, ) There is a paper for failure types for web applications. Comparison between failure types of web applications and web services would be a good starting point Different types of modeling: Colored Petri net, Markov Chain, or Path based modeling Model distributed composite web services