[Users] iops limiting

Mark Johanson mjohanson at a2hosting.com
Tue Feb 11 14:53:20 PST 2014


So I'm still trying to figure out this iops limiting for vm's and it 
still doesn't seem to be doing much. I have dropped the limit to a iop 
limit of 50 and the results were not much different then 1000 or 500 limits:

> THETEST ->       IOPS    Transfer
> 4096 direct:            104210   427   MB/s
> 4096 dsync:             39   160   kB/s
> 4096 sync:              29   119   kB/s
> 4096 fdatasync:         56549   232   MB/s
> 65536 direct:           6309   414   MB/s
> 65536 dsync:            12   834   kB/s
> 65536 sync:             12   829   kB/s
> 65536 fdatasync:         3734   245   MB/s
> 1048576 direct:         403   424   MB/s
> 1048576 dsync:          9   10.3   MB/s
> 1048576 sync:           9   9.7   MB/s
> 1048576 fdatasync:         232   244   MB/s
>
> Running IOPing I/O benchmark...
> IOping ->                       IOPS    Transfer
> Direct Small IO
> Direct Large IO
>
> Serverbear IOping Tests         IOPS    Transfer
> IOPing Default                  96      0.4
> IOPing seek rate:               136     0.5
> IOPing sequential:              404     101.1
> IOPing cached:                  323257  1262.7

It took 254m18.987s (according to "time ./script" to finish running).  
I'm being asked by management on why iop numbers are not changing when a 
vm is limited and I have no answers. Am I doing something wrong? 
Shouldn't a iop limit of 50 be a drastically different number then 500?

On 01/22/2014 03:07 PM, Mark Johanson wrote:
> I'm not sure if I am doing something wrong (which is very possible) 
> but I am not seeing any kind of significant change in numbers when 
> going from a iops unlimited setting on a vps to a iops 500 limit on 
> the same vps. The vm is set via:
>
> vzctl set 101 --iopslimit 1000 --save
> rebooted and cat'd the conf file does show that the 500 limit is there:
>
> IOPSLIMIT="500"
>
> vzctl version 4.6.1
> 2.6.32-042stab084.14 #1 SMP Mon Dec 16 18:16:56 MSK 2013 x86_64 x86_64 
> x86_64 GNU/Linux
>
> For example the following comes from a variety of tests running on the 
> disks:
>
> _Unlimited test_
>
> Downloading FIO...
> Downloading IOPing...
> Downloading UnixBench...
> Creating a file to use.
>
> THETEST ->       IOPS    Transfer
> 4096 direct:            108881   446   MB/s
> 4096 dsync:             31   129   kB/s
> 4096 sync:              20   85.3   kB/s
> 4096 fdatasync:         56781   233   MB/s
> 65536 direct:           6235   409   MB/s
> 65536 dsync:            21   1.4   MB/s
> 65536 sync:             21   1.4   MB/s
> 65536 fdatasync:         3969   260   MB/s
> 1048576 direct:         414   435   MB/s
> 1048576 dsync:          15   16.1   MB/s
> 1048576 sync:           15   16.0   MB/s
> 1048576 fdatasync:         231   243   MB/s
>
> Running IOPing I/O benchmark...
> IOping ->                       IOPS    Transfer
> Direct Small IO
> Direct Large IO
>
> Serverbear IOping Tests         IOPS    Transfer
> IOPing Default                  93      0.4
> IOPing seek rate:               192     0.7
> IOPing sequential:              945     236.2
> IOPing cached:                  325603  1271.9
> Time,Hostname,Comment,4096 direct-iops,4096 direct-rate,4096 direct-unit,4096 dsync-iops,4096 dsync-rate,4096 dsync-unit,4096 sync-iops,4096 sync-rate,4096 sync-unit,4096 fdatasync-iops,4096 fdatasync-rate,4096 fdatasync-unit,65536 direct-iops,65536 direct-rate,65536 direct-unit,65536 dsync-iops,65536 dsync-rate,65536 dsync-unit,65536 sync-iops,65536 sync-rate,65536 sync-unit,65536 fdatasync-iops,65536 fdatasync-rate,65536 fdatasync-unit,1048576 direct-iops,1048576 direct-rate,1048576 direct-unit,1048576 dsync-iops,1048576 dsync-rate,1048576 dsync-unit,1048576 sync-iops,1048576 sync-rate,1048576 sync-unit,1048576 fdatasync-iops,1048576 fdatasync-rate,1048576 fdatasync-unit,IOp 4k D-iops:IOp 4k D-megs,IOp 64k D-iops,IOp 64k D-megs,SB Default-iops,SB Default-megs,SB Seek rate-iops,SB Seek rate-megs,SB Sequential-iops,SB Sequential-megs,SB Cached-iops,SB Cached-megs,
> 1390349785,server.example.com,,108881,446,MB/s,31,129,kB/s,20,85.3,kB/s,56781,233,MB/s,6235,409,MB/s,21,1.4,MB/s,21,1.4,MB/s,3969,260,MB/s,414,435,MB/s,15,16.1,MB/s,15,16.0,MB/s,231,243,MB/s,,,,,93,0.4,192,0.7,945,236.2325603,1271.9,
>
> _500 Limited Test_
>
> THETEST ->       IOPS    Transfer
> 4096 direct:            95545   391   MB/s
> 4096 dsync:             40   164   kB/s
> 4096 sync:              39   163   kB/s
> 4096 fdatasync:         61423   252   MB/s
> 65536 direct:           6676   438   MB/s
> 65536 dsync:            40   2.7   MB/s
> 65536 sync:             40   2.6   MB/s
> 65536 fdatasync:         3852   252   MB/s
> 1048576 direct:         402   422   MB/s
> 1048576 dsync:          27   28.8   MB/s
> 1048576 sync:           27   28.5   MB/s
> 1048576 fdatasync:         238   250   MB/s
>
> Running IOPing I/O benchmark...
> IOping ->                       IOPS    Transfer
> Direct Small IO
> Direct Large IO
>
> Serverbear IOping Tests         IOPS    Transfer
> IOPing Default                  117     0.5
> IOPing seek rate:               196     0.8
> IOPing sequential:              805     201.2
> IOPing cached:                  325274  1270.6
> Time,Hostname,Comment,4096 direct-iops,4096 direct-rate,4096 direct-unit,4096 dsync-iops,4096 dsync-rate,4096 dsync-unit,4096 sync-iops,4096 sync-rate,4096 sync-unit,4096 fdatasync-iops,4096 fdatasync-rate,4096 fdatasync-unit,65536 direct-iops,65536 direct-rate,65536 direct-unit,65536 dsync-iops,65536 dsync-rate,65536 dsync-unit,65536 sync-iops,65536 sync-rate,65536 sync-unit,65536 fdatasync-iops,65536 fdatasync-rate,65536 fdatasync-unit,1048576 direct-iops,1048576 direct-rate,1048576 direct-unit,1048576 dsync-iops,1048576 dsync-rate,1048576 dsync-unit,1048576 sync-iops,1048576 sync-rate,1048576 sync-unit,1048576 fdatasync-iops,1048576 fdatasync-rate,1048576 fdatasync-unit,IOp 4k D-iops:IOp 4k D-megs,IOp 64k D-iops,IOp 64k D-megs,SB Default-iops,SB Default-megs,SB Seek rate-iops,SB Seek rate-megs,SB Sequential-iops,SB Sequential-megs,SB Cached-iops,SB Cached-megs,
> 1390359348,server.example.com,,95545,391,MB/s,40,164,kB/s,39,163,kB/s,61423,252,MB/s,6676,438,MB/s,40,2.7,MB/s,40,2.6,MB/s,3852,252,MB/s,402,422,MB/s,27,28.8,MB/s,27,28.5,MB/s,238,250,MB/s,,,,,117,0.5,196,0.8,805,201.2325274,1270.6,
> Does there need to be a specific setting enabled on the node, specific 
> type of scheduler, etc that is not being set and why I am not seeing 
> any significant changes? Any help would be great.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvz.org/pipermail/users/attachments/20140211/a1a7344b/attachment.html>


More information about the Users mailing list