An administrator observes that virtual machine storage activity on an ESXi 6.x host is negatively affecting virtual machine storage activity on another host that is
accessing the same VMFS Datastore.
Which action would mitigate the issue?
A.
Enable Storage IO Control.
B.
Configure Storage DRS.
C.
Enable the Dynamic Queue Depth Throttling option.
D.
Configure the Disk.SchedNumReqOutstanding parameter.
Explanation:
Explanation/Reference:
A is correct.
URL: https://pubs.vmware.com/vsphere-60/index.jsp?topic=%2Fcom.vmware.vsphere.resmgmt.doc%2FGUID-7686FEC3-1FAC-4DA7-B698-B808C44E5E96.html
See also https://blogs.vmware.com/vsphere/2014/05/enabling-monitoring-storage-io-control.html
At a basic level SIOC is monitoring the end to end latency of a datastore. When there is congestion (the latency is higher then the configured value) SIOC reduces the latency by throttling back VM’s who are using excessive I/O. Now you might say, I need that VM to have all of those I/O’s, which in many cases is true, you simply need to give the VMDK(s) of that VM a higher share value. SIOC will use the share values assigned to the VM’s VMDK’s to prioritize access to the datastore.
Just simply turning SIOC on will guarantee each VMDK has equal access to the datastore, shares fine tune that giving you the ability to give VMDK’s more or less priority during times of contention.
Why wouldn’t the answer be Storage DRS? Is there only one LUN? In my mind a good administrator would use Storage DRS to eliminate hot spots across multiple LUNs. In my mind Storage IO Control doesn’t solve the problem it just limits the bandwidth to that hot spot. Read this and tell me what you think?
http://www.yellow-bricks.com/2012/05/22/an-introduction-to-storage-drs/
The question says – accessing the same VMFS Datastore
Yes it does, Storage DRS would move one of those VM’s so it is not on the same VMFS LUN and that would resolve the issue. Does that make sense or am I missing something?
In general, you Matthew are correct as such approach is more fundamental to resolving disk I/O performance issues. But that assumes there’s a datastore (DS) cluster with SDRS enabled / configured. This however is not what is in the question, at least not explicitly. So I guess whoever at VMware came up with this question meant the least expensive, right-at-hand type of a solution, which in this case is use of Storage I/O.
SDRS works based on I/O latency and space utilization thresholds. Poor performance is relative and maybe below the set thresholds of SDRS in the environment and so won’t be triggered.
Also you may not have anywhere to move the VM due to lack of free space yet you can utilize SIOC to resolve this issue.
Matt – SDRS is one of the options, but in production systems you cannot generate additional load by enabling SDRS and move machine to different DS. So, better option would be to enable SIOC.