Towards a fast multi-tier storage system simulator

This paper presents StorageSim, a multi-tier storage system simulator. StorageSim is a process-based discrete-event simulator developed using the SimPy simulation framework. It simulates the operation of a multi-tiered storage system; for example, a system that stores super “hot” files in non-volatile RAM, less “hot” files in solid state drives (SSDs) and “warm” and “cold” files in hard disk drives (HDDs). StorageSim comes with three data-placement policies, and can be extended to support other policies. It can replay publicly available storage traces from the Storage Networking Industry Association (SNIA) and other public sources, and can be used to evaluate data-placement policies prior to implementing them on a real system. By abstracting away many complex details, StorageSim provides a fast simulation framework that can be used to simulate large scale storage systems. Experimental results show that StorageSim is useful, can reproduce prior results from real deployments, and is fast enough to handle Big Data workloads in a timely manner.