Dependable Grid Services: A Case Study with OGSA-DAI.

Grid middleware usually makes use of several software modules that due to their complexity and development approach may have some latent bugs and leaks. These bugs can cause visible performance failures and undesired service crashes. To cope with this sort of transient failures we present a proactive software rejuvenation approach that exploits the use of virtualization middleware. To prove the effectiveness of our mechanism we decided to apply it to OGSA-DAI, a sound example of a middleware that has been widely used in several Grid-related projects. OGSA-DAI makes use of Tomcat/Axis as the SOAP container and Axis v1.2.1 suffers from memory leaks. When it is not configured properly these leaks will result in a crash of the OGSA-DAI server. In this paper, we explain the application of our rejuvenation scheme in this particular example and we show that it is easy to get a software-based approach to improve the availability of a Grid Service even when one of the underlying layers suffers from clear symptoms of software aging. Our ultimate goal is to give a contribution for the techniques and concepts that can be used to achieve dependable Grid services. [ABSTRACT FROM AUTHOR]