Benchmarking Persistent Memory with Respect to Performance and Programmability

Abstract

High performance computing is to process data and do calculation on that data at high speed. Traditionally high performance computing uses dynamic random access memory in their computers. This thesis will explore an alternative form of memory called persistent memory that will occupy half the DIMM slots in the computer. The persistent memory will be tested for its performance and will be compared to traditional memory. The performance will also be tested in competition and cooperation with traditional memory and its performance will be evaluated. This thesis will also have a look at how to program with persistent memory, are there any programming choices that could be recommended.