Comparative Study Between in-Memory Database and Hard Drive Database: An Application to Data Analytics

Selecting a database implementation fit for specific data analytic purposes is often challenging due to manufacturers' bias in database performance reporting. This paper aims to compare disk-based and in-memory database implementations to assist businesses in selecting the best database for their database use case. The paper utilised a comparative methodology to evaluate the performance of disk-based versus in-memory databases using TPC-C benchmarks. Our analysis revealed that diskbased databases consistently outperformed in-memory for the Online Transaction Processing workload. The result was due to the table lock performed by the IMDB on every Data Manipulation Language statement, which allows only one session to write to the table at a time. The findings have practical implications for businesses, namely, the advantage of disk-based databases for online transaction processing workloads, thus presenting cost savings. The recommendation has positive economic implications as disk-based databases require less capital investment than in-memory databases.