Fast Peak-to-Peak Restart for SSD Buffer Pool Extension

A promising usage of Flash SSDs in a DBMS is to use it to extend the main memory buffer pool by caching in the SSD selected pages that are evicted from the buffer pool. These schemes have been shown to produce big performance gains in the steady state. Simple methods for using the SSD buffer pool throw away the data in the SSD when the system is restarted (either when recovering from a crash or restarting after a shutdown), and consequently need a long “ramp-up” period to regain peak performance. A recent method to address this limitation is to use a memory-mapped file to store the metadata (called the SSD buffer table) about the contents of the SSD buffer pool, and to recover the metadata at the beginning of recovery. However, this method can result in lower sustained performance, because every update to the SSD buffer table may incur a random I/O operation. In this paper we propose two new designs. One design reconstructs the SSD buffer table using transactional logs. The other design asynchronously flushes the SSD buffer table, and upon restart, lazily verifies the integrity of the data cached in the SSD buffer pool. We have implemented the previously proposed scheme and these two new schemes in SQL Server. For each design, both the write-through and the writeback caching policies were implemented. Using two OLTP benchmarks (TPC-C and TPC-E), our experimental results show that our designs produce up to 3.8X speedup on the interval between peak-to-peak performance, with negligible performance loss; in contrast, the previous approach has a similar speedup but up to 54% performance loss.