V + H: Hybrid architecture for DSS and OLTP

Abstract

Previous research has been done to identify the best pragmatic approach that DBMSs must follow to manage different workloads. Some approaches have followed the “One size fits all” approach, trying to incorporate all features in a row-oriented DBMS (also called horizontal) to manage both OLTP and DSS workloads. Additionally, there have been specialized DBMS, following a columnar approach (also called vertical), that focuses on the growing demand to efficiently manage DSS workloads. This paper proposes a combination of both vertical and horizontal DBMS to best manage OLTP and DSS workloads. We have used mature, commercially available products from a single vendor and developed a custom middleware Decision Query Module that identifies the best option for most efficient execution of a query. This V+H architecture also offers the functionality of a mirrored DB without paying twice for storage.