Seeking Fast, Durable Data Management: A Database System and Persistent Storage Benchmark
In-memory database systems (IMDSs) eliminate much of the performance latency associated with traditional on-disk database management systems (DBMSs), but some applications require a higher level of data durability (i.e. an ability to recover if someone “pulls the plug” or otherwise disrupts volatile memory). As a solution, IMDSs offer transaction logging, in which changes to the database are recorded on persistent media. But critics object that logging re-introduces the storage-related latency that builds slowness into on-disk DBMSs.
Will an IMDS with transaction logging outperform a traditional DBMS? Will type of storage – hard disk drive vs. SSD vs. state-of-the-art memory-tier device – affect any performance difference?
McObject’s benchmark report, In Search of Data Durability and High Performance: Benchmarking In-Memory & On-Disk Databases With Hard-Disk, SSD and Memory-Tier NAND Flash answers these questions with tests using IMDS and on-disk DBMS technology, along with commercially available hardware including Fusion ioDrive2 NAND flash memory from Fusion-io. It also examines the impact of transaction length (long vs. short) on database system performance and the interplay of storage device type with this effect.