vSphere Storage Appliance

 View Only
  • 1.  Common causes for bad disk performance?

    Posted Aug 11, 2010 10:54 PM

    It is often mentioned that one of the most common bottlenecks in a vSphere system is disk I/O. I am wondering if there is anything general that could be said about this, as in what causes this?

    Is it the hard drives in an array that has too slow RPM, too long seek time, the SAN that has a too small cache (or incorrectly configured) or the wrong RAID type for the wrong type of application? Or too many VMs on the same datastore? Or too much load on the Storage Processor? Or the FC fabric overloaded or the bottleneck of 1 Gb ethernet on iSCSI? Or something else?

    I do of course know that this can never be answered with a simple answer that will fit all situations. However, I am very curious of your experience. When you have troubleshooted bad disk IO, were there any typical things that were wrong?



  • 2.  RE: Common causes for bad disk performance?

    Posted Aug 11, 2010 11:15 PM

    You already mentioned most issues.

    One to add: Local disks - even 15k SAS disks - on a RAID controller without BBWC/BBU for operating in write-back mode.

    André



  • 3.  RE: Common causes for bad disk performance?

    Posted Aug 12, 2010 05:59 AM

    You already mentioned most issues.

    But of these, are there one or more that is more common a problem than others?



  • 4.  RE: Common causes for bad disk performance?

    Posted Aug 12, 2010 01:17 AM

    In my experience, i had issues with too slow disks (SATA 7200 RPM). This works fine with some VMs which don't need high performance, like dev/QA environment.

    However, if your VMs need high performance like database servers, the disk speed it's a big concern. But this is just one among many concerns about the storage performance.

    The cache size is also a big deal. If your SAN has a large cache and this is well configured, then the RAID configuration isn't so important. Otherwise, the RAID type would become in a critical aspect in order to get a good performance of your VMs.

    The number of VMs on one datastore is important too. It's recommended to distribute the VMs on smaller datastores instead of group all the VMs on just one big datastore. This improve the load balancing of storage resources.

    There are too many factors when you are troubleshooting a disk performance issue. All depend of the VMware infrastructure you are troubleshooting, the Storage device, and how you connect the ESX/ESXi hosts with the Storage device.




    Regards / Saludos

    -


    Patricio Cerda !http://www.images.wisestamp.com/linkedin.png!

    VMware VCP-410

    Join to Virtualizacion en Español group in Likedin

    See My Blog

    See My Linkedin Profile

    -


    Si encuentras que esta o cualquier otra respuesta ha sido de utilidad, vótalas. Gracias.

    If you find this or any other answer useful please consider awarding points by marking the answer helpful or correct. Thank you.



  • 5.  RE: Common causes for bad disk performance?

    Posted Aug 16, 2010 12:03 PM

    The cache size is also a big deal. If your SAN has a large cache and this is well configured, then the RAID configuration isn't so important

    Is there some recommended cache size or some best practice how the read / write should be balanced in a typical situation?