|
|
|
Forum Newbie
Group: Forum Members
Last Login: Wednesday, June 25, 2008 4:11 PM
Posts: 7,
Visits: 17
|
|
I've been running IOMeter in a VM that gets it's storage from a VSA..
While the test is running I'm noticing the VSA is using 90% ++ CPU.
Can I help performance if I add a second vCPU?
Also with 7 drives in a RAID 5 configuration.
Do you have any performance bench marks to compare against?
TIA
|
|
|
|
|
Supreme Being
Group: Moderators
Last Login: Yesterday @ 2:22 PM
Posts: 101,
Visits: 354
|
|
I don't have performance numbers I can compare against that. Most of the time I see clusters of 2 or 3 VSA doing replication.
That said, I'd guess you should be getting around 90 MB/s and 700-800 IOPS.
As for CPU running at even 100% is not a problem. We mean for nodes to be well utilized.
You can add more CPU and it might improve performance a bit.
I'd not expect much main line performance improvement though. I'd expect it to affect other operations, like how fast a snapshot deletes.
Adam C
Product Manager
LeftHand Networks
|
|
|
|
|
Forum Newbie
Group: Forum Members
Last Login: Wednesday, June 25, 2008 4:11 PM
Posts: 7,
Visits: 17
|
|
Can you say how (which tool) you are measuring that performance?
IOMETER?
Can you list the test parameters so we have something to test against to make sure our VSAs are configured correctly?
# of outstanding I/Os?
Access Specifications? Bytes, % of Read, % of Random?
Thanks
|
|
|
|
|
Junior Member
Group: Forum Members
Last Login: Sunday, July 20, 2008 4:31 PM
Posts: 20,
Visits: 105
|
|
Hi Marc
Firstly you might be able to help us out with a bit more information, in order to help you.
Your Hardware down to model, make and chipsets if you can, CPU , Memory, RAID card etc.
Also VMs on that machine and the VSA machine.
Vkernel (free 15 day trial) or Platespin can help analyse the ESX, if you do not have a VC.
Process Explorer from Sysinternals for the VM and maybe NewtPro to show the Page file usage on these VMs. (Not the VSA).
Also you might want to read, the VSA could be paging cache to disk without knowing it.
http://vsaforum.lefthandnetworks.com/Topic33-13-1.aspx
Hope this helps
Reality is defined by the physical universe and your mind - "Its the second one that varies" :-)
|
|
|
|
|
Forum Newbie
Group: Forum Members
Last Login: Wednesday, June 25, 2008 4:11 PM
Posts: 7,
Visits: 17
|
|
Thanks for the suggestions:
1 See my reply to the cross posting you listed.
2 My VSA results
VSA Configuration
2 - Dell 2950 - 2-E5320 CPUs-Quad Core 1.86Ghz - 8GB RAM - Perc 5i
RAID0 - 36GB - ESX installed here no VMFS's
RAID5 - 7 Drives - 146GB SAS 10K 3Gbps Drives 2.5"
VMFS is 4MB Block Size - 816GB space on RAID5 - Exclusive to VSAs
2 VSAs with 2-way replication 7.0.01.6046
Each VSA has 1024MB RAM and a reservation of that amount for RAM.
CPU reservation of 1800Mhz
Host running IOMeter is a VM W2K Pro SP4
IOMeter Results
Drive is on VSA - through ESX - E: Drive - VM on Remote ESX Host - iSCSI target is remote
512B; 50% Read; 100% Random, 32 I/O = IOs/Sec=1242; Tot MB/Sec=3.03; Avg IO Resp=25.74; Max IO Resp=2733; CPU %=9.76
64K; 100% Read; 0% Random, 32 I/O = IOs/Sec=1280.97; Tot MB/Sec=80.06; Avg IO Resp=24.98; Max IO Resp=1613; CPU %=14.25
64K; 0% Read; 0% Random, 32 I/O = IOs/Sec=714.4; Tot MB/Sec=44.65; Avg IO Resp=44.76; Max IO Resp=1549.13; CPU %=7.04
Drive is on Local VMFS - through ESX - Accessing via Physical Drive no Drive Letter
512B; 50% Read; 100% Random, 32 I/O = IOs/Sec=8249; Tot MB/Sec=4.03; Avg IO Resp=3.88; Max IO Resp=2350; CPU %=37.38
64K; 100% Read; 0% Random, 32 I/O = IOs/Sec=1258; Tot MB/Sec=78.61; Avg IO Resp=25.44; Max IO Resp=154; CPU %=6.58
64K; 0% Read; 0% Random, 32 I/O = IOs/Sec=376; Tot MB/Sec=23.5; Avg IO Resp=85.08; Max IO Resp=331.44; CPU %=4.23
Drive is on Local VMFS - through ESX - Accessing via Drive Letter:
512B; 50% Read; 100% Random, 32 I/O = IOs/Sec=8247; Tot MB/Sec=4.03; Avg IO Resp=3.88; Max IO Resp=146; CPU %=44.20
64K; 100% Read; 0% Random, 32 I/O = IOs/Sec=1250; Tot MB/Sec=78.14; Avg IO Resp=25.59; Max IO Resp=167; CPU %=8.10
64K; 0% Read; 0% Random, 32 I/O = IOs/Sec=390; Tot MB/Sec=24.39; Avg IO Resp=82.00; Max IO Resp=535.65; CPU %=4.39
Drive is on VSA - through ESX - E: Drive - VM and iSCSI target on same system - Using Internal vSwitch
512B; 50% Read; 100% Random, 32 I/O = IOs/Sec=2004.37; Tot MB/Sec=.98; Avg IO Resp=16.96; Max IO Resp=2288.54; CPU %=14.78
64K; 100% Read; 0% Random, 32 I/O = IOs/Sec=933.22; Tot MB/Sec=58.33; Avg IO Resp=34.29; Max IO Resp=1875.70; CPU %=10.93
64K; 0% Read; 0% Random, 32 I/O = IOs/Sec=967.7; Tot MB/Sec=60.48; Avg IO Resp=33.06; Max IO Resp=3890.15; CPU %=10.96
Drive is on VSA (With 2CPUs) - through ESX - E: Drive - VM and iSCSI target on same system - Using Internal vSwitch
512B; 50% Read; 100% Random, 32 I/O = IOs/Sec=3380.81; Tot MB/Sec=1.65; Avg IO Resp=9.46; Max IO Resp=1432.90; CPU %=25.68
64K; 100% Read; 0% Random, 32 I/O = IOs/Sec=1019.48; Tot MB/Sec=63.72; Avg IO Resp=31.39; Max IO Resp=1455.36; CPU %=12.22
64K; 0% Read; 0% Random, 32 I/O = IOs/Sec=1026.16; Tot MB/Sec=64.13; Avg IO Resp=31.18; Max IO Resp=1450.08; CPU %=11.74
Also posted as an attachment.
|
|
|
|