Expected Speeds

Software-based VM-centric and flash-friendly VM storage + free version

Moderators: anton (staff), art (staff), Max (staff), Anatoly (staff)

CCSNET_Steve
Posts: 16
Joined: Fri Dec 20, 2013 9:51 pm

Sat Jan 25, 2014 3:57 pm

Anatoly (staff) wrote:Two questions for now:
*Can you confirm that you have used GPT formatting instead of MBR on every layel that is involved into the test?
*Can I ask if you have some additional Windows Server box that you can use to run the IOmeter (that should exclude the problems in the virtualization layer) ?
Thank you
Hi,

no currently everything is at MBR, i didn't see anything about it having to be GPT and being that the partitions are relitivly small i didn't see the need to go to GPT, also the virtual machines (that have been around a while) are using MBR.

the "physical server" in the image above is an "additional windows box" that i have connect via 4Gb links

later today i'm increasing the 4 x 1Gb links to 8 x 1Gb links, would the recommendation to multipath all 8 on separate IPs or bond 4 x 2 ??

regards
CCSNET_Steve
Posts: 16
Joined: Fri Dec 20, 2013 9:51 pm

Sat Jan 25, 2014 7:19 pm

is GPT that much of a factor, as the disk speed on the starwind box, and the physical host is as expected

regards
User avatar
Anatoly (staff)
Staff
Posts: 1675
Joined: Tue Mar 01, 2011 8:28 am
Contact:

Mon Jan 27, 2014 9:52 am

is GPT that much of a factor, as the disk speed on the starwind box, and the physical host is as expected
The disk aligns is one of the most common reasons of the low performance issues (the other one is outdated driver), and the GPT helps to elliminate that problem, so we are always recommend to use it.
later today i'm increasing the 4 x 1Gb links to 8 x 1Gb links
Never use NIC teaming, just stick with MPIO Round Robin.
Best regards,
Anatoly Vilchinsky
Global Engineering and Support Manager
www.starwind.com
av@starwind.com
CCSNET_Steve
Posts: 16
Joined: Fri Dec 20, 2013 9:51 pm

Mon Jan 27, 2014 10:16 am

I have been looking at the disk alignment recently but i havn't been able to find a suitable tool / test to figure out if the discs are correctly aligned do you have any resources (starwind or otherwise) that you could suggest to work through, i believe i have to check from raw stripe, through starwind host, iscsi presentation, ESX lun and finally the virtual host windows, is this correct ?

and reference the MPIO that was my plan just wanted to ensure it was correct

regards
User avatar
Anatoly (staff)
Staff
Posts: 1675
Joined: Tue Mar 01, 2011 8:28 am
Contact:

Mon Jan 27, 2014 5:31 pm

Unfortunatelly we don`t have one, but I`ve found the doc that should give you the basic idea:
http://technet.microsoft.com/en-us/libr ... .100).aspx

And I want to remind you that I strongly recommend to use GPT instead of MBR, which should eliminate disk aligns issues.
Best regards,
Anatoly Vilchinsky
Global Engineering and Support Manager
www.starwind.com
av@starwind.com
CCSNET_Steve
Posts: 16
Joined: Fri Dec 20, 2013 9:51 pm

Mon Jan 27, 2014 5:37 pm

Anatoly (staff) wrote:Unfortunatelly we don`t have one, but I`ve found the doc that should give you the basic idea:
http://technet.microsoft.com/en-us/libr ... .100).aspx

And I want to remind you that I strongly recommend to use GPT instead of MBR, which should eliminate disk aligns issues.
GPT on what the host ?
User avatar
Anatoly (staff)
Staff
Posts: 1675
Joined: Tue Mar 01, 2011 8:28 am
Contact:

Thu Jan 30, 2014 11:14 am

Actually everywhere where you have NTFS - on the volume where the Starwind images are stored, on the VM, etc.
Best regards,
Anatoly Vilchinsky
Global Engineering and Support Manager
www.starwind.com
av@starwind.com
CCSNET_Steve
Posts: 16
Joined: Fri Dec 20, 2013 9:51 pm

Thu Jan 30, 2014 6:34 pm

Thanks,

but as we can't easily change / test this i'm unsure to the next step, it seems to be an issue between VMFS and your system the network has tested fine, the local disks test fine, if anyone is reading this software and using it with ESX/VMFS i would be greatful to hear from you, as currently we cannot justify a purchase when we simply can't get the throughput via our VMWare infrastructure

regards
User avatar
Anatoly (staff)
Staff
Posts: 1675
Joined: Tue Mar 01, 2011 8:28 am
Contact:

Thu Jan 30, 2014 7:43 pm

We can verify that. Can you create the RAM disk, connect to it directly from VM and run the benchmark?
Best regards,
Anatoly Vilchinsky
Global Engineering and Support Manager
www.starwind.com
av@starwind.com
CCSNET_Steve
Posts: 16
Joined: Fri Dec 20, 2013 9:51 pm

Thu Jan 30, 2014 8:31 pm

do mean
1. to create a ramdisk, mount it inside ESX, write VMFS partition, mount inside a VM and test the speed
or
2. mount a ramdisk as an iscsi volume on a server hosted within our ESX ?

i have done the former, and posted the results further up this thread, i can't do the later as the storage lan is seperate from the VM LAN and i'd have to make changes to achieve this

regards
User avatar
Anatoly (staff)
Staff
Posts: 1675
Joined: Tue Mar 01, 2011 8:28 am
Contact:

Fri Jan 31, 2014 9:57 am

I meant option #2 - that fully excludes the physical storage part, and leaves only network and StarWind.
Best regards,
Anatoly Vilchinsky
Global Engineering and Support Manager
www.starwind.com
av@starwind.com
CCSNET_Steve
Posts: 16
Joined: Fri Dec 20, 2013 9:51 pm

Fri Jan 31, 2014 10:03 am

how would you expect results other than what we recieve when connecting a physical host to the storage network ?

regards
User avatar
Anatoly (staff)
Staff
Posts: 1675
Joined: Tue Mar 01, 2011 8:28 am
Contact:

Mon Feb 03, 2014 10:03 am

In the test that I`ve suggested you hsould use the RAM disk, which is the virtual disks that is based on the random access memory. So basically this excludes the storage part from the test entirely, so as all hte problems that may be related to this, such as disk aligns, hardware malfunction, some missconfiguration, etc.
Best regards,
Anatoly Vilchinsky
Global Engineering and Support Manager
www.starwind.com
av@starwind.com
Post Reply