ACTDP: An Adaptive Chunk Tool for Database Partition in Shared-nothing Distributed Database

摘要

Shared-nothing, distributed database is the core component for modern OLTP application. It uses horizontal partitioning to partition data into multiple partitions to achieves both ACID and scalability. The existing distributed transactions degrade the performance of database, many partition management systems therefore had been proposed to use graph partitioning to mitigate impact of distributed transactions on performance. These systems run well when amount of data is limited, but consume a lot of resources otherwise. This paper proposed an adaptive chunk tool ACTDP, automatically aggregate the vertices in the graph in such systems by consider both number and weight of each vertex, improves performance of these graph-based partition management systems.