Tag: disk latency

Disk defragmenting SharePoint on Virtual Machines: Performance tweak or myth?

I previously posted about the effectiveness of defragmenting guest machines (SharePoint servers) in virtualized environments here:
SharePoint on Virtual Machines – Is Disk Defragmenting necessary for Performance?

Today the ever-helpful President of PerfectDisk, Bob Nolan, wrote me to share a very interesting tidbit that he found on VMware’s VSphere documentation:

“Last week I was researching something and I found an interesting piece of documentation that might be helpful to you down the road.  The VMware 5.1 doc has a section enumerating 12 things to do when you have disk latency or I/O contention issues.  “Defragment the file system of all the guests” is the 2nd recommendation on the list http://pubs.vmware.com/vsphere-51/index.jsp?topic=%2Fcom.vmware.vsphere.monitoring.doc%2FGUID-174326D5-238B-48CA-B030-02009E388523.html



“You may want to look at this too. Cormac Hogan is a VMware storage architect and this is a blog he did on how Storage I/O Control (SIOC) balances fairness and performance http://blogs.vmware.com/vsphere/2013/04/virtual-machine-io-fairness-versus-performance.html .  The gist of the blog is that VMware Kernel settings reduce the number of outstanding I/O requests any VM can have when multiple VMs are sharing a LUN.  SIOC basically throttles performance to improve latency.  Further, if you do sequential I/O, VMware will grant you more I/O requests.

In response I wrote this paper which says if VMware is going limit your I/O requests then get the most from the ones you have by doing larger I/O. When you do larger I/O you tend to do sequential I/O so you can also get the additional requests VMware is willing to grant http://www.raxco.com/user_data/white_papers/Getting_More_VMware_Performance_from_Fewer_IO_Requests.pdf

This reinforces the original conclusion: This data combined with my decade-plus experience of seeing direct noticeable performance improvements on IIS boxes after full defragmentation and implementation of regular defrags, located on physical or virtual disks, leads to me to stand by my guns: I will continue to recommend disk defragging.  Would love to see if someone can change my mind..