Hi Guys,
so i have a question regarding a feature called Auto Tiering on a Compellent storage device in our special use case and i hope that somebody has some experience or can give me a punch in the right direction.
I work for a software development company with an i would say medium sized virtualized infrastructure. we host around 420 VM´s, about 90% of them we need for the software development process. So developers run compile / build jobs etc. on them. The workload we see is very asymetric, because those jobs don´t run on every vm on every time. Some vm´s have load only 1 time in a week some two times and so on. But if they do have load, it is very i/o heavy. There are a lot of (thousands) of small files that the compilers will read and after compilation objects will also be written to the disk. So in burst scenarios we see around 20k-30k Iops with a 60%:40%, sometimes 50%:50% read/write distribution.
At present time we have a netapp storage infrastructure with FC on a FAS3140 HA Cluster (15KSAS disks all). The infrastructure is around 4 years old and we need to change it. We have problems performance wise and also capacity wise. We are now looking at a Compellent as a competitor to a refreshed Netapp infrastructure. In the presentation DELL showed us a feature called Auto Tiering which made me very suspicious. In their example they had 3 tiers with T1 with SSD´s, T2 with 10K SAS disks and T3 with high capacity SATA drives. I have big concerns that this feature is contraproductive in the workload scenario we have. For example if a VM is on T1, performance will be great, now the VM will not be used, let´s say for a week (which is quite common in the development process). Because of that it will go down to T3 on the SATA drives and the next time the build job runs, everything will be read from the slow drives and Performance might be worse than it is now.
Has anybody any experience with this in the real world, thoughts etc ? Everything will be highly appreciated.
Thanks