Possibilities and Limitations of using Cloud Object Stores for a ”One Size Fits All” Database

Abstract

Having a single database system serving both OLAP and OLTP use cases is highly desired, yet has never been achieved. Simultaneously, object stores along with metadata storage layers emerged as a new technology, unifying Data Lake and Data Warehousing capabilities. This work examines the feasibility to also unify OLAP and OLTP requirements with this technology. This was done by implementing a novel database system called "deltadb," leveraging object stores. This system was evaluated with a performance benchmark and its results can make us conclude that it is possible to build a system that can serve both OLAP and OLTP workloads. However, the performance on that use cases still lacks those of applications that were particularly designed for a specific use case. Overall, the usage of such a system can only be recommended in very specialized situations.