Squeezing the Last Drop for VDI and HPC – Nutanix’s Replication Factor

So I apologize in advance to the Nutanix Support team. The beauty of Nutanix with its Google like file system, NDFS(Nutanix Distributed File System) is the ease of use with performance. A distributed architecture that scales and massive parallelism at 25,000 IOPS a block is a lot to handle but some people are just plain greedy. If you see the work that Jim Moyle has done with Windows 7, you realize a Win 7 desktop will eat as many IOPS it can get it’s hands on.

Nutanix has a setting called Replication factor(RF). Replication factor controls the number of copies of data within a container. These copies exist for fault tolerance: if a physical disk fails, the cluster can recover data from another copy cleanly. The cluster manages the replicated data, and the copies are not visible to the user. The Replication factor cannot be changed from the GUI, it’s defaulted to 2. If you want to change this setting you need to do it from Nutanix Command Line interface(NCLI). You’re able to get the NCLI as a download from the management console. I am not sure of the command yet but when I do I will add it to my resource section of this blog.

Changing this setting should be done with extreme caution as your DBA’s would freak as would I if I didn’t have it set to 2. Other workloads like temp space, swap and linked clones for VDI can be a great use case. Your VDI pools should be sized so they aren’t taking three hours to do a recompose task and I would say the vast majority of VDI uses cases could live within the SLA. HPC software usually has some from working directory that is just used for running a job that would benefit. If your HPC environment is responsible translating decisions into money, the risk is well worth the reward.

Look for more on this topic in the future.

Speak Your Mind