Web Service Cost OptimizationHussain AljaferDepartment of ComputerScienceWayne State UniversityDetroit, Michigan, USAhussain.aljafer@wayne.eduKhayyam HashmiDepartment of ComputerScienceWayne State UniversityDetroit, Michigan, USAkhayyam@wayne.eduZaki MalikDepartment of ComputerScienceWayne State UniversityDetroit, Michigan, USAzaki@wayne.eduAbdelkarim ErradiDepartment of ComputerScience and EngineeringQatar UniversityQatarerradi@qu.edu.qaABSTRACTCloud computing has become a major building block forWeb applications where the service providers provide thecostumers with Web services performing certain tasks sothe system developers can use these services instead of im-plementing them. Service computing follows the “pay-per-use” as a model for pricing. Due to the variety of ser-vice providers, we have many dif f erent prices for servicesand these services have dif f erent performance characteris-tics. The problem here is how to f i nd the best services interms of cost and performance and is it feasible to switchfrom one service provider to another. In this paper we pro-pose Wcost, a scheme to resolve this problem based on P-OCEA algorithm which is a combination of P-Optimalityand Genetic algorithms which result in more ef f i cient so-lution than Genetic Algorithm approaches. Our proposedscheme includes a f l exible and easily customizable objectivefunction that is suitable for dif f erent types of Web servicesto be composed.1. INTRODUCTIONDue to the rapid increase and spread of Web enabled com-puting many service providers compete to of f er better ser-vices to the users. The provided quality of the of f ered servicedif f ers form one service user to another according to the rel-ative importance of each feature provided by the service. Tojudge the service quality i.e. to assess which service may bebetter for a certain user in a certain task, several param-eters are put to consideration. Finding the cheapest ratedoes not mean this service is the best service since thereare other parameters judging this. For example in videostreaming services, the video resolution is a primary featureof the service. In this case, having a service with a very lowcost and a lower resolution than what the application needsis not an ef f ective solution. In some cases, the user alreadyhave a service that performs the desired tasks but want someimprovement to be added to the application. The improve-ment could be in terms of performance or operational cost.The problem here is to f i nd another service such that it isfeasible to switch from the current service provider to thenew one.Minimizing the cost to the user does not mean f i nding thecheapest service since most of the time the cheapest serviceis not the most ef f i cient. There is usually a relation be-tween cost and performance and in most cases the cheaperthe cost leads to lower performance measures. those perfor-mance measures include Availability, Reliability, ResponseTime, Cost and in some cases some other domain specif i cmeasures depending on the type of application. Thereforeour main problem becomes“ Given a relative importance foreach measure, f i nd the best candidate serviceˆ a ? A ´ Zâ ? A ´ Z. Thisincludes the case whether the user has no current servicefor the task or even if the user has a service but looking forfurther enhancement. In other words, is it feasible and costef f ective for the user to switch from one service provider toanother service provider and if it is then which one is thebest solution. Our proposed scheme uses the P-OCEA algo-rithm proposed in [1] to f i nd the solution to optimize the costin Web service composition according to a given objectivefunction. The P-OCEA algorithm is faster in f i nding thesolution than Genetic algorithms. The Algorithm uses theP-Optimality Criteria combined with Genetic Algorithm’soperations to f i nd the solution. P-Optimality plays the roleof candidate selection for solution comparison according tothe provided objective function instead of calculating a f i t-ness value for each candidate which will enhance the perfor-mance of f i nding a solution to the problem.The rest of the paper is divided as follows:Section IIcontains a brief discussion of Genetic Algorithms, SectionIII describes P-Optimality, Section IV Describes P-OCEA.Section V discusses the derivation of our objective functionderivation and how it can be customized. We discuss somerelated works in Section VI, Section VII shows the exper-iments we conducted along with experimental results and- 116 -Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for prof i t or commercial advantage and that copies bear this notice and the full citation on the fi rst page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specif i c permission and/or a fee.MEDES '15, October 25-29, 2015, Caraguatatuba, Brazil© 2015 ACM. ISBN 978-1-4503-3480-8/15/10…$15.00http://dx.doi.org/10.1145/2857218.2857237
