Having just completed a storage/ VMware migration from 4 x Dell Equallogics to a single Nimble CS300 (48TB 2.4TB Cache 1GB iSCSI) I just wanted to share with you my experience.
The firm runs a number of highly transactional Microsoft SQL servers which accepts data from the likes of Compare the Market and GoCompare. They have approximately 75 vm’s and also a small VDI environment running on the Nimble. Unsurprisingly the Nimble handles all this without breaking in to a sweat, but that’s to be expected right ?
So why do we get so much performance from a 3U box with 12 SATA disks and 4 SSD’s; quite simply CASL
CASL is a CPU-driven storage architecture, so write performance is no longer dependent on spindle speed or spindle count. What’s more, CASL dramatically increases the usable capacity of disk and flash through compression and its unique sequential data layout.
This document provides a great overview on CASL for those that want to know more.
So after the environment had bedded in I took a look at the Nimble interface to see what the steady state IOP and latency figures were for this environment, bearing in mind we have several highly transactional SQL servers, 75 vm’s and a small virtual desktop environment all running on this box, lets take a look below.
IOP figures for an average working day
- Random Read: 881 IOPS
- Sequential Read: 1931 IOPS
- Random Write: 580 IOPS
- Sequential Write: 180 IOPS
Latency figures for an average working day ( 1GB iSCSI Network)
- Read Latency 0.42msec
- Write Latency 0.19msec
Pretty impressive figures considering the blended workload, this is about the 11th Nimble I have installed and I am always impressed with the performance of these arrays.