A critical element to every customer engagement is a detailed analysis of that user's environment by a Compellent Storage Architect. One part of that analysis is a detailed examination of the I/O workload over a 24 hour period. That workload becomes the basis for determining the quantity of tier 1 (enterprise) disk to sustain that user's workload plus a reasonable margin. Since Compellent makes all tier 1 disk and their aggregated performance available to all servers, the actual disk utilization is much higher than those environments where each physical disk is allocated to only a specific server. In those environments when the corresponding server switches between varying periods of activity and inactivity, the associated disks married to that server are idling while efficiency is being lost forever.

An optional feature called "Fast Track" through software optimizes disk performance my moving the most frequently accessed blocks to the outer cylinders of each drive. This causes the shortest possible seeks between frequently used data. While this technique could be thought of as short stroking, the entire disk is capable of holding data with the innermost cyclinders contain the inactive data blocks. This technique in extreme conditions can double the I/O performance of a disk drive while substantially reducing latency. We think of this as yet another additional green feature which can limit the number of tier 1 disks required to maintain performance.

The remaining required storage capacity is typically implemented with tier 3 disk. Since 80% of the data in a volume is inactive, this becomes an excellent choice for housing that data. In the future, when it is necessary to add more disk capacity, historical performance information is utilized to determine the type of tier to augment. If performance levels are great, additional tier 3 drives are added. Conversely if performance needs to be augmented, additional tier 1 drives are added.

First of all, if you can get a SAN just for ESX , that is the best course of action. Having said this, the most important thing to remember is that ESX and the VM don't write to the SAN simultaneously; this data is held and written in pulses, then the ESX host is ready--this what some may call dirty writes. Also the more drives spread across the RAID, the faster things will be. If possible, you will also want to increase the write cash on you SAN to combat all the dirty writes ESX will be making. Also remember the cost of Fibre drives over SATA drives; this is offset by the speed of SATA and dictates which RAID type you will end up using.

For example:
If you use 8 fibre drives, it is possible to configure RAID5 with no problems (in most cases), but if you use 8 SATA drives you may have to do RAID0+1 (10) to get the same performance. Also remember most companies have a limited budget, so configure wisely.