Cost and Time Prediction for Efficient Execution of Bioinformatics Workflows in Federated Cloud

Abstract

Cloud computing has devised an interesting computational model which provides a set of features such as storage, database and processing power, all made available as services. Recently, the concept of cloud computing has been extended to federated cloud computing in which different providers interconnect to provide more resources in an integrated and transparent way to the end user. Thus, the use of cloud platforms has been widely encouraged in applications that require a lot of processing and/or storage power, such as workflows in Bioinformatics. Users who operate such workflows are faced with a very large variety and amount of available resources, making it difficult to choose the correct ones for a certain workflow. This measurement is far from trivial and, in order to address this problem, this paper proposes an approach called sPCR (Cost Prediction and Computational Resources Service), which mixes GRASP metaheuristics and the multiple linear regression method with the purpose of dimensioning the resources to the users in a transparent way. In addition, sPCR allows the user to interact and choose between high-performance, low-budget runs, or set how much to pay and how long to finish workflows, all automatically and transparently. The results show that sPCR is able to efficiently estimate the resources, costs and execution time of workflows.