What does it mean when someone say that their VM is having 8 * 500 IOPS?


  • What does it mean when someone say that their VM is having 8 * 500 IOPS?

    I am having D12_V2 size VM machines. Analysis team analyzed the disks and gave below feedback

    1. H: drive: Average Seconds Per transfer is 1.791
    2. H: has up to 104meg/second in reads, and up to 13 meg/second in writes; IOPS is low, no more than 520 transfer per second. 
    3. O: drive is also doing poorly: average Seconds per Transfer is .048(=48ms)  
    4. O: has up to 85 meg/second in writes, very few reads (this is normal for a transaction log disk!) and only 520 IOPS. 

    Does it mean I need to go for high IOPS machines? 

    Many Thanks Deepak

    Wednesday, July 6, 2016 9:47 AM

All replies

  • IOPS (input/output operations per second) is performance measurement used to benchmark computer storage devices. 

    The Standard_D12_v2 has a max of 8x500 IOPS. If you require more IOPS you can look at using Premium Storage. Premium Storage supports DS-series, DSv2-series and GS-series Azure Virtual Machines (VMs)

    You can see the IOPS sizes here for your current The Standard_D12_v2 or the ones listed above that support Premium Storage.

    You can read more about Premium Storage here

    I would show those links to your Analysis team and work out what size VM's you need to suit their disk performance.

    If this post was helpful, please click the up arrow or propose as answer.
    Nick |

    Wednesday, July 6, 2016 10:10 AM
  • Greetings Deepak,

    IOPS is a measure, created to relate machines to one another.
    Ofcourse depending on the workload, your milage may differ and it's fairly hard to say based on your feedback that you require a bigger environment.

    but i guess the Analysis Team was expecting better performance from the disks; which might be related to knowledge of similar processes within your company.

    if they feel this way(and pay) i wouldn't be scared to upgrade the machine.
    However if it's comming out of the pocket, i would keep measures on Queue-length and IO-waites to determine that the workload occures as the Analysis-team expected.

    With kind regards,

    Wednesday, July 6, 2016 11:00 AM
  • So, shall i go for disk pooling (disk stripes) ? or shall I upgrade to premium ?

    Many Thanks Deepak

    Thursday, July 7, 2016 10:53 AM
  • I would ask your analysis team as they are the ones performing the tests. Show them the Premium storage documentation and check with them. They know more about the environment. 

    Nick |

    Thursday, July 7, 2016 11:02 AM