Research On Multi-Granularity Evolution Of Web Services Composition Based On QoS

With the style of service computing (SOC), a variety of heterogeneous resourcesdistributed on the Internet can be encapsulated into services published for others toshare. Web services technology conforms this style, Through the combination of Webservices could integrate resources efficiently, improve the reuse of resources. At thesame time, give a convenience to large-scale programming.On the other hand, open, dynamic environment of the Internet, and changing userneeds make the web service composition relative to the traditional softwaredevelopment has its inherent characteristics (such as cooperativity, evolvability, etc.).The breakdown of environment of web service and congestion of the networkenvironment could may be severely degraded the web service QoS, In order to makethe service composition instance can continue to meet the needs of people, a certainevolution of service combination is necessary. The existing research for compositeservice evolution mostly concern the correctness of control flow and data flow afterevolution as well as instances of online migration, a small amount considering theevolution of the global QoS research mostly only for a single service, for the lack ofexisting research, propose the problem of multi-granularity evolution in this paperand research to this, the main work is as follows:(1) Propose the concept of multi-granularity evolution of QoS-based web servicecomposition. Generated for a single abstract service class candidate services toreplace has two drawbacks in combination model:1.Not necessarily be able to findthe right services in the candidate service to replace the original service, such ascandidate services QoS does not satisfy the global QoS constraints.2.Only for theabstract service to generate candidate service, not necessarily be able to get theoptimal global QoS. Based on the above two considerations, we propose whenevolution of service, generated the replacement area of larger particle size of servicefor single service is about to replaced in service combination model. Replacementareas of each of the different particle size can also generate candidates, this expandedsearch space of candidate service and conducive to further obtain better solutions tomeet global QoS constraints.(2) Given the algorithm of generating all the replacement area to a service node.The premise of generating a larger particle size of alternative service is generatinglarger particle of replacement area in combination model, In this regard, we propose pre-existing control region of a control node and to the single service needs to bereplaced,take a self-bottom up to the direction of the input and output of the servicegradually extended to obtained a replaceable service area including this services, onthis account get a combination of multi-granularity model.(3) For different granularity process model in evolutionary combination model,using improved discrete particle swarm algorithm to find global QoS optimal value onglobal QoS constraints and propose a new optimal evaluation model, this model takesinto account the combination of model granularity. Discrete particle swarm algorithmhas the common defect with the particle swarm algorithm, which is easy to fall intothe local optimum in the late of algorithm. To this defect, in order to ensure thediversity of particles in late of algorithm, we learn from the idea of hybridizationvariation in the genetic algorithm and select a certain of particles to hybridization.experiments show that it make for particle swarm algorithm get better global optimalvalue.