Abstract

Cloud computing has emerged as a new computing model that enables scientists to benefit from several distributed resources such as hardware and software. Clouds are an opportunity for scientists that need high performance computing infrastructure to execute their scientific experiments. Most of the experiments modeled as scientific workflows manage the execution of several activities and produce a large amount of data. Since scientists work with large amounts of data, parallel techniques are often a key factor on the experimentation process. However, parallelizing a scientific workflow in the cloud environment is far from trivial. One of the complex tasks is to define the configuration of the environment, i.e. the number and types of virtual machines and to design the parallel execution strategy. Due to the number of options for configuring an environment it is a hard task to be done manually and it may produce negative impact on performance. This paper proposes a lightweight cost model that is based on concepts of quality of service (QoS) in cloud environments to help determining an adequate configuration of the environment according to restrictions imposed by scientists.