PTL: Partitioned Logging for Database Storage on Flash Solid State Drives

We propose Partitioned Logging (PTL), a storage layout for databases on flash solid state drives. In PTL, we replace data writes with logging, and put data and logs into separate blocks. Moreover, we group data blocks into partitions so that updates on each partition are appended as log entries to one log block. This way, we can tune the partition size to balance the read and write performance based on the hardware and workload characteristics. We have implemented PTL in PostgreSQL, which involves moderate changes to the buffer manager, the storage manager, and the transaction manager. We have also developed an analytical model to determine the PTL parameter values. We have evaluated PTL using the standard TPC-C benchmark as well as homegrown workloads. The empirical results match our analytical analysis, and show a considerable improvement over both the traditional storage and a leading flash-based database storage scheme.