HiPAS: High Performance Adaptive Schema Migration - Evaluation of a Self-Optimizing Database Migration

摘要

HiPAS is a database migration method, aimed at reducing downtime during offline migrations by automatically adapting to available system resources. Investigating the applicability of adaptive capabilities for database migrations, two stages of system complexity, adaption and anticipation, were mapped onto the requirement of utilizing a system up to an optimal degree in order to achieve the shortest possible transfer duration. The developed method is automated by implementing the HiPAS software, which adapts to its environment by continuously monitoring relevant system information, and increasing or decreasing the current parallelization degree whenever necessity is assumed. To enable a flexible adaption, the total amount of migration data is partitioned into equal sized transfer jobs being distributed across available instances and networks. Since HiPAS is invoked on the database layer, and controlled by a temporarily created autonomous database user, migration metadata is stored inside tables thus being highly integrated with the actual migration data. HiPAS was designed and evaluated iteratively following the IS research framework and reveals significant downtime reduction potential compared to non-adaptive migration approaches like Oracle "Data Pump". Our results serve as a contribution to all researchers and practitioners in investigating fields of application for adaptability mechanisms.