FlashLogging: exploiting flash devices for synchronous logging performance

• Authors: Shimin Chen
• Venue: Proceedings of the 2009 ACM SIGMOD International Conference on Management of data
• Year: 2009

Synchronous transactional logging is the central mechanism for ensuring data persistency and recoverability in database systems. Unfortunately, magnetic disks are ill-suited for the small sequential write pattern of synchronous logging. Alternative solutions (e.g., backup servers or sophisticated battery-backed write caches in high-end disk arrays) are either expensive or complicated. In this paper, we exploit flash devices for synchronous logging based on the observation that flash devices support small sequential writes well. Comparing a wide variety of flash devices, we find that USB flash drives are a good match for this task because of its unique characteristics: widely available USB ports, hot-plug capability useful for coping with flash wear, and low price so that multiple drives are affordable. We propose FlashLogging, a logging solution that exploits multiple (USB) flash drives for synchronous logging. We identify and address four challenges: (i) efficiently exploiting multiple flash drives for logging; (ii) coping with the large variance of write latencies because of device erasure operations; (iii) efficient recovery processing; and (iv) combining flash drives and disks for better logging and recovery performance. We implemented our solution within MySQL-InnoDB. Our real machine experiments running online transaction processing workloads (TPCC) show that FlashLogging achieves up to 5.7X improvements over magnetic-disk-based logging, and obtains up to 98.6% of the ideal performance. We further compare our design with one that uses Solid-State Drives (SSDs), and find that although SSDs improve logging performance, multiple USB flash drives can achieve comparable or better performance with much lower price.