I spend a lot of time writing blog posts but I also spend time with customers and their VDI projects. During my time with customers I have identified that read/write ratios in VDI environments could be anything from 50/50 to 20/80 in favor of writes.
It turns out that every once in a while someone comes to me saying that a given VDI deployment have more read IOs than write IOs. Even if Windows guest image is optimized to extremes and applications virtualization is utilized I would still say VDI has more writes than reads. Well, am I wrong?
Why is that important?
This is extremely important when sizing the storage array that is going to support the VDI environment. Read my article Get hold of VDI IOPs, Read/Write Ratios and Storage Tiering to understand how the RAID type will determine the numbers of spindles required to support the workload based on the amount of IOPS and Read/Write ratio.
I would like to validate numbers with REAL Production data from various VDI deployments. With the results in hands I will publish an article that will help everybody to understand this skew a little better.
So, fill the form bellow with data from your organization. The data need to be REAL and the name of the company and email address is required for validation purposes. I will keep the data provided in absolute confidentiality.
The instructions about how to collect the data are bellow.
When the 100th survey is answered I’ll give away a USD $50 Giftcard (any country) for a lucky winner.
- Select the following day of a normal operations day for your VDI environment (for most organizations Tuesday trough Thursday)
- Select a host in use by your VDI environment and go to the Advanced Performance Chart.
- Select Datastore > Custom…, then select your organizations peak production time (for most organizations 8:00AM trough 5:00PM), then in Objects select active datastores supporting VDI VMs.
- On Counters, select Average read requests per second and Average write requests per second.
- Note: Take a screenshot of the configuration and the performance chart for validation, if required.
Below are the Maximum and Average numbers we want to consolidate.
Thanks for supporting this research!