Adaptive Query Processing in Data Grids

Abstract

The data grid integrates wide-area autonomous data sources and provides users with a unified data query and processing infrastructure. Adaptive data query and processing is required by data grids to provide better quality of services (QoS) to users and applications in spite of dynamically changing resources and environments. Existing AQP techniques can only meet partially data grid requirements. Some existing work is either addressing domain-specific or single-node query processing problems. Data grids provide new mechanisms for monitoring and discovering data and resources in a cross-domain wide area. Data query in grids can benefit from this information and provide better adaptability to the dynamic nature of the grid environment. In this work, an adaptive controller is proposed that dynamically adjusts resource shares to multiple data query requests in order to meet a specified level of service differentiation. The controller parameters are automatically tuned at runtime based on a predefined cost function and an online learning method. Simulation results show that our controller can meet given QoS differentiation targets and adapt to dynamic system resources among multiple data query processing requests while total demand from users and applications exceeds system capability.