MacroDB: Scaling Database Engines on Multicores?

Abstract. Multicore processors are available for over a decade, but gen-eral purpose database management systems (DBMS) still cannot fully explore the computational resources of these platforms. This paper ex-plores a simple and easy to deploy approach for improving DBMS perfor-mance in multicore platforms, by maintaining multiple database engines running in parallel, rather than a single instance, thus circumventing the increase in contention due to load interactions. Unlike previous works, we focus on in-memory DBMS, exploring different design solutions that combine distributed systems and concurrent programming techniques. We show that we are able to improve performance over standalone solu-tions, without modifying either database or application code, by up to 3 times while minimizing response times.