Title: DERIVATION OF STOCHASTIC REWARD NET (SRN) FROM UML SPECIFICATION CONSIDERING COST EFFICIENT DEPLOYMENT MANAGEMENT OF COLLABORATIVE SERVICE COMPONENTS

Issue Number: Vol. 1, No. 3
Year of Publication: Oct - 2011
Page Numbers: 721-742
Authors: Razib Hayat Khan, Poul E. Heegaard
Journal Name: International Journal of New Computer Architectures and their Applications (IJNCAA)
- Hong Kong

Abstract:


Performance evaluation of distributed system is always an intricate undertaking where system behavior is distributed among several components those are physically distributed. Bearing this concept in mind, we delineate a performance modeling framework for a distributed system that proposes a transformation process from high level UML notation to SRN model and solves the model for relevant performance metrics. To capture the system dynamics through our proposed framework we outline a specification style that focuses on UML collaboration and activity as reusable specification building blocks, while deployment diagram identify the physical components of the system and the assignment of software artifacts to identified system components. Optimal deployment mapping of software artifacts on the available physical resources of the system is investigated by deriving the cost function. The way to deal with parallel thread processing of the network nodes by defining the upper bound is precisely mentioned to generate the SRN model. The proposed performance modeling framework provides transformation rules of UML elements into corresponding SRN representations and also the prediction result of a system such as throughput. The applicability of our proposed framework is demonstrated in the context of performance modeling of a distributed system.