Quality of Service support for Grid Storage Environments.
Vijay Velusamy, Anthony Skjellum · GCA · 2006
Grid computing enables the utilization of geographically distributed heterogeneous resources for processing data. In such large-scale applications, availability of data is essential for efficient utilization of the resources, especially since data utilized by the applications may not be available at the site of execution. The distributed nature of these grid applications and the disparate nature of storage resources warrant some guarantees to access the data that may be distributed among grid locations, while considering economic and utility values. Quality of Service (QoS) serves to allow delivery of data in real-time to enable data consuming applications to execute without awaiting completion of data generating applications. Though there have been studies to analyze data streaming in grid computing environments, these strategies have minimal support for any type of QoS or utility value. This paper presents a novel approach to support QoS requirements for grid storage systems, by offering statistical QoS assurances using data replication, rather than a best-effort basis. A framework for QoS enabled data-intensive applications in a grid environment based on pNFS is designed for this purpose.