With the rapid development of hardware, a stand-alone computer can employ a memory which has large amounts of volumes. Several industries and research institutions have devoted more resources to develop several in-memory databases, which preload the data into memory for further processing. With the boom of in-memory databases, there emerges requirements to evaluate and compare the performance of these systems impartially and objectively. In this paper, we proposed MemTest, a novel benchmark considering the main characteristics of an in-memory database. This benchmark constructs particular metrics, which cover CPU usage, cache miss, compression ratio, minimal memory space and response time of an in-memory database and are also the core of our benchmark. We design a data model based on inter-bank transaction applications, around which a data generator is devised to support the data distributions of uniform and skew. The MemTest workload includes a set of queries and transactions against the metrics and data model. In the end, we illustrate the efficacy of MemTest through implementations on three different in-memory databases.
展开▼
