PM aware storage engine for MongoDB

Abstract

With the maturity of Persistant Memories (PM) such as storage class memory technologies, e.g., STT-MRAM, PCM, ReRAM and 3DXpoint, we expect to see practical implementation of data structures, data stores and databases for use-cases such as IoT, mobile, and cloud. We developed a PM-aware storage engine for MongoDB which leverages PM hardware capabilities such as byte addressability and persistency. With our storage engine we see improved latency, less write amplification, less capacity and simpler implementation due to the fact that some code paths become unnecessary compared to past implementations.