An Advanced Caching Solution to Cluster Storage Environment

Abstract

Clustered storage is the deployment of multiple data servers working together to improve reliability, capacity and performance. Clustering divides workloads to every storage server to control and monitor workload transfer and file access between servers without taking into account of the physical location of the file. Solid State Drives (SSD) can be considered as a more sophisticated version of a USB memory stick since the memory stick does not have any moving part associated with it and moreover, data is stored in microchips. In this paper, we give an overview of an advanced caching solution to improve IO and application performance by using flash storage in cluster storage environment. It is a cluster storage solution with two highly scalable servers with optimizations to ensure fast service failovers and deploying one or two solid state drives as the cache devices for faster and better performance. The software supports write-back caching policy where both read and write requests on hot regions of drivesare cached. With write-back, write requests to the hot regions are acknowledged immediately after it is written to the cache device and this (dirty) data will be flushed to back-end virtual drive in the background. Flushing of dirty data will be performed by the flush manager of the software under different scenarios like amount dirty data reaches a threshold, IO activity during a time interval is low etc. The solution effectively harnesses the flash storage performance potential by retaining only frequently accessed data in flash for quick retrieval. The solution provides unmatched efficiency, performance, support and reliability for enterprises or storage world.