|Home \ Graduation Activities \ Post-Graduation Page||Login|
Instrumentation, Profiling and Auditing Distributed Web-Service Transactions in SOA N-Tier Platforms
The existing tools and techniques for profiling, instrumentation and optimization of SOAs and EIA platforms and AppServer technology are not general enough to track, audit and characterize the flows of internal data operations and transactions in multi-tier distributed applications (or, for ex., in architectures inspired in a reference model like SEDA). For example, the information, code-generation and patterns observed from profilers and test&auditing generated components only capture localized or limited patterns or partial behaviours of long and distributed multi-layering asynchronous transactions (ex., IBM WebSphere , BEA WebLogics). Other existing tools and profiling techniques and models for this purpose from recent research contributions (Whodunit, MagPie, DS-BlackBoxes, are also not enough to capture end-to-end behaviours for instrumentation, profiling and testing multi-layering distributed WS composable transactions. Long and asynchronous distributed transactional profiling and instrumentation in NTier software architectures is hard, by many factors: (i) difficulties in anticipating real-runtime conditions or real execution traces, (ii) distribution patterns and dynamics of concurrent execution flows, (iii) conjugation in single mechanisms and generic algorithms of different synchronization and information passing patterns, (iv) asynchronous event-based executions, etc,. At the same time, the interferences among concurrent transactions and shared processing flows and the impact of profiling and instrumentation mechanisms are additional complex problems to deal with. In this thesis the objectives will be: (1) an extensive and detailed study-survey of profiling, instrumentation and test-auditing and test-generation problems and techniques, particularly appropriate to capture long multi-layering asynchronous transactional flows in the context of SOA and/or EIA platforms; (2) a comparative study on related work approaches from two/three representative industrial development environments and the specific solutions coming from the recent research proposed techniques, to deal with the above difficulties with more generic solutions (the dissertation will contribute with a critical discussion about the argued generality of such proposed solutions, recognizing the limitations for the case of SOA and EIA environments) and (3) a proposal and implementation of new mechanisms and algorithms, or the possible adaptation of others, recently proposed, for the scenario of WS long asynchronous composable transactions. The proposed mechanisms and algorithms must be able to not require specific support at the OS level or any specific knowledge about each application behaviour. The solution must be particularly focused on instrumentation techniques, profiling and test&auditing mechanisms and code generation facilities for a JAVA Based Application Server SOA and for the case of WS composable multi-tier asynchronous long transaction flows. The proposal will be complemented and validated by the use of the proposed algorithms and mechanisms in a (demo/or/real) large scale NTier application (to be used as a reference workbench / benchmarking). The validation must propose metrics and show measurements to antecipate the impact study of (a) the overhead that those mechanisms can introduce in terms of performance (response-times), (b) the impact on operations throughput, (c) load overheads in the different distributed tiers and (d) specific execution times of different algorithms or variants. If possible, the implementation can be inspired by the practical development and integration of profiling and auditing plugins in a specific programming environment (ex., the Eclipse Platform).
Start Date: 2007-11-01
End Date: 2008-10-31
Post-Graduation Student / Researcher / Professor: