Repozytorium publikacji - Politechnika Gdańska

In Service Oriented Architecture, applications are developed by integration of existing services in order to reduce development cost and time. The approach, however, requires algorithms that select appropriate services out of available, alternative ones. The selection process may consider both optimalization requirements, such as maximalization of performance, and constraint requirements, such minimal security or maximum development cost. The selection problem has NP-hard computational complexity in a general case, which causes the necessity to use approximate algorithms. The paper compares existing algorithms and identifies typical algorithm characteristics. The algorithms are based on two main processing models: graph model using multiconstraint optimal path selection and combinatorial model using 0-1 multidimensional multichoice knapsack problem. Computational complexity and accuracy were considered in the analysis using both theoretical calculations and known experimental results. Algorithms use typically some simplifications during processing, which may cover either filtering of potential solutions or a simplified application model. Additionally, specific features of algorithms were identified, including: dynamic adaptation of service selection during execution and application of artificial intelligence methods.

Autorzy

• dr inż. Paweł Kaczmarek link otwiera się w nowej karcie