So the VMware guys ran a test to find out how many IOPs can one of the ESX systems can push. The article below talks about a test which was performed and proved that one ESX machine, given a specific workload, can sustain over 100K IOPS.
http://blogs.vmware.com/performance/2008/05/100000-io-opera.html
Workload characteristics are: 100/50/8 (100% random, 50% reads (and 50% writes) and 8KB block size)
From the storage side, it took 3 CX3-80s with 495 15K RPM disk drives to push this workload (Wow!). Considering that there is nothing else running on that CX3-80 than this workload and the test was stopped at this point because the storage arrays were saturated. The articles goes into some tests to prove this.
So a CX3-80 with 11 trays of FC disk drives (total of 165 disk drives) can deliver up to 103671/3 = 34557 IOPs. That’s It ?? That is less than 210 IOP/s per FC 15K Disk Drive!!
I took a quick look at the SPC benchmarks for other technologies. Looking at the NetApp FAS3170 results, on the summary page, it lists the SPC-1 IOPs at 60,515.34. The system tested contained 16 shelves of 14 drives each, bring to a total of 224 drives (about 270 IOPs per FC drive). 2 of these FAS3170s will more than likely will deliver what 3 CX3-80s could not..
Wrong answer. A drive is a drive is a drive. 10k drives do 140 to 150 iop/s, 15k drives do between 180 and 190 iop/s. Thats it. There is no 270 iop/s from any single drive being deliverd at the fibre channel block level. The Netapps box doesn’t present block mode storage, it presents file systems. Period. Understand ?? It is irrelevant whether the spindles are running in an HDS 9980, HP EVA, EMC Symmetrix or Netapps Filer. ******************** [lousy language deleted]
@Saldarismg
You are correct in saying that a drive is a drive is a drive. I guess what I meant to drive at in my post was the effective IOPs per drive as a metric varies from storage system to storage system (eIOPS as opposed to IOPS, I agree). Various manufacturers employ different technologies to use those drives in a different fashion. The NetApp storage systems (filers, as folks call it) does use a file system called ONTAP. CLARiiON uses FLARE Operating System and Symmetrix uses DART. They all use the underlying disks differently to deliver what is needed to the customer.
cheers
We are investigating to run Vmware esx on a server connected to HDS9980 storage and wyse thin client.The most important problem is the voice delivery for Call Center.RDP is handling one way voice and should be disabled,we have to activate sound card + usb and head set:
we need the 1)advise on the best performance configuration of server,storage,san switch and ext. 2)advise on equipment for voice delivery.
If there is a best practice already in operation would be appritiated.
thanks
nouri
Understood Saldarismg, that you don’t know anything about SANs & NASs.
Rajeev is not talking about local storage my friend , he is talking about SAN&NAS. Also NetApp has a block level presentation of data. Please read before posting.