Research on manufacturing grid resource service optimal-selection and composition framework

In order to address the resource service optimal-selection (RSOS) and composition problem in manufacturing grid (MGrid) system and provide high-quality service to users, an MGrid RSOS and composition framework (MGrid-RSOSCF) is investigated in this study. The process of RSOS and composition is divided into the following five steps in MGrid-RSOSCF: (1) decomposing the submitted manufacturing task into several subtasks (i.e. single resource service requested task) if the submitted task is a multiple resource service requested task; (2) searching out the qualified resource service for each decomposed subtask and generating the corresponding candidate resource service set; (3) retrieving, evaluating and comparing the quality of service (QoS) for each candidate resource service, and provide data for service optimal-selection and composition –if the submitted task is a single resource service requested task; (4) evaluating synthetically the overall quality of each candidate resource service and ranking them, and selecting the optimal one for the task – if the submitted manufacturing task is an multiple resource service requested task; (5) selecting one candidate resource service from each candidate resource service set and constructing a new composite resource service according to the submitted task requirements, and collecting all the possible resource service composite execution paths (RSCEP) and selecting the optimal paths to execute the task. The proposed MGrid-RSOSCF consists of five layers and each layer provides the corresponding necessary services and algorithms to address one problem mentioned above. The five layers are: (1) T-layer, responsible for MGrid task decomposition; (2) S-layer, responsible for resource service match and search; (3) Q-layer, responsible for QoS processing; (4) O-layer, responsible for evaluating and ranking the candidate resource service and (5) C-layer is responsible for resource service composition and optimal-selection. The case study and comparison of performances of the algorithms demonstrate that the proposed methods are sound on success rate and executing efficiency.