Improving OLAM with cloud elasticity

Abstract

Elasticity is considered one of the fundamental properties of cloud computing, and can be seen as the ability of a system to increase or decrease the computing resources allocated in a dynamic and on demand way. This feature is suitable for dynamic applications, whose resources requirements cannot be determined exactly in advance, either due to changes in runtime requirements or in application structure. A good candidate for using cloud elasticity is the Ocean-Land-Atmosphere Model (OLAM), since it presents a significant load variation during its execution and due to online mesh refinement (OMR), that causes load unbalancing problems. In this paper, we present our efforts to adapt OLAM to use the elasticity offered in cloud environments to dynamic allocate resources according to the demands of each execution phase, and to minimize the load unbalancing caused by OMR. The results show that elasticity was successfully used to provide these features, improving the OLAM performance and providing a better use of resources. © 2014 Springer International Publishing.