Infrasrtucture update - TS-459U

[lmassimo]

Hi to all and sorry for my english,

I manage about 12VM in a cluster with:

2 x Host Esx 4.1.0.260247 on HP DL380G5 (single cpu and 16GB of RAM for any host)
1 x QNAP TS-459U

Both Esx use QNAP NFS share as shared storage.

All work fine but

- Some time vm user (terminal services) report some "tears" of some seconds. Terminal session "freeze" for some second and then restart. Different time in a day. Users in session in different VM report some issue.

- Snapshot usage for some VM is impracticable. If i delete a shapshot all VM in same node become unusable utill delete operation is completed.

We think that QNAP is in overload e we would upgrade it with new and more performance NAS. We think to QNAP TS-809U to use with iSCSI instead of NFS.

Is this a good point of start?

I can view "differences" whit this new NAS?

I hope to be clear.

Thanks in advanced for any help.

Massimo.

[schumaku]

Give you slightly more time - and consider the TS-879 Pro, TS-1079 Pro (ox enclosures) or the TS-1079U and TS-1279U in 19" enclosures becoming availane somewhat later. http://qnap.com/PressRelease_detail.asp?pr_id=241
I would not convert to a TS-809 anymore - unless there is fire under the roof.

[lmassimo]

Hi shumacu and thanks for your reply.

I have read article at link suggested, but i not found information about any date. Accordly with you i not change with TS-809 but how long will I wait for new models?

In any case, about "technical side" of the question? Upgrading storage is the best way to solve my problem?

Thanks.

Regards.

Massimo.

[PiroNet]

Hi to all and sorry for my english,

I manage about 12VM in a cluster with:

2 x Host Esx 4.1.0.260247 on HP DL380G5 (single cpu and 16GB of RAM for any host)
1 x QNAP TS-459U

Both Esx use QNAP NFS share as shared storage.

All work fine but

- Some time vm user (terminal services) report some "tears" of some seconds. Terminal session "freeze" for some second and then restart. Different time in a day. Users in session in different VM report some issue.

- Snapshot usage for some VM is impracticable. If i delete a shapshot all VM in same node become unusable utill delete operation is completed.

We think that QNAP is in overload e we would upgrade it with new and more performance NAS. We think to QNAP TS-809U to use with iSCSI instead of NFS.

Is this a good point of start?

I can view "differences" whit this new NAS?

I hope to be clear.

Thanks in advanced for any help.

Massimo.

NFS is a good fit for your environment, don't change it.
You haven't told us the number of disks, vendor/model and RAID config you run on your 459U
You may be interested in a free tool called VKernel StorageVIEW to identify datastore bottlenecks...

[lmassimo]

Hi Pironet and thanks for your reply.

I have 3 disks WDC WD2003FYYS-02W0B01.0 configured in RAID 5. Is not good?

I look immediatly for tool you have suggested.

Thanks.

Massimo.

[PiroNet]

I have 3 disks WDC WD2003FYYS-02W0B01.0 configured in RAID 5. Is not good?

You have made a good choice there. These disk are Raid Edition and therefore fully compatible with TLER.
No you will have to look at the VMs and see which one generates high IOPS. Once you have found it you will have to fix it or throttle it to ensure it doesn't hog all your storage IOPS.

FYI in your current disk setup, high IOPS is anything bigger than 300 IOPS on a sustain rate.
You can have burst peaks higher that 300 but on a very short period of time (>20ms). Most of the time caches and queues will absorb and mitigate those bursts...
You may enable write cache for EXT4 but you should then install an UPS to protect your servers/storage devices.

[lmassimo]

No you will have to look at the VMs and see which one generates high IOPS

I can use tool that you have suggested me or i can have another way to find this info?

You can have burst peaks higher that 300 but on a very short period of time (>20ms).

Sorry, did you min <20ms ? Or i have not undestrand...

I have found suggestion about the number of HDD. I found suggestion about splitting (obviously in another storage such as TS-1079 Pro) in 8 or 10 HDD of (for instance) 600GB configured in RAID 5 can increase considerably the performance. With your knowledge can you confirm this? Making this type of upgrade i have a justified benefit?

Thanks another time for your help.

Regards.

Massimo.

[PiroNet]

Answers in bold

No you will have to look at the VMs and see which one generates high IOPS

I can use tool that you have suggested me or i can have another way to find this info?
The tool can help you but there is no 'real' history, so you can go back and see the last 4 hours for example...
Use the vCenter Client to connect to either a vCenter server (if you have one) or directly to the ESX server and then look at the performance tab.

You can have burst peaks higher that 300 but on a very short period of time (>20ms).

Sorry, did you min <20ms ? Or i have not undestrand...
Oops, yes I meant <20ms

I have found suggestion about the number of HDD. I found suggestion about splitting (obviously in another storage such as TS-1079 Pro) in 8 or 10 HDD of (for instance) 600GB configured in RAID 5 can increase considerably the performance. With your knowledge can you confirm this? Making this type of upgrade i have a justified benefit?
More you add spindles (disks) more you have IOPS to distribute among your virtual machines.
You could also go for SSD disks if the total IOPS required by your 12 virtual machines exceed by far 300 IOPS.
Or you throttle down the virtual machine that generates too much IOPS. Read more at Limiting Disk I/O From A Specific Virtual Machine


Thanks another time for your help.

Regards.

Massimo.