When VDI first started coming out many storage companies were concerned about the effect boot storms, shutdown storms, AV scan storms, etc. would have on system performance. As such, they were keen to demonstrate how well their systems did against boot storms and other un-natural IO activity to support VDI environments.
But at last years SFD4 and during subsequent discussions around the storage round table noone mention boot storms as being a concern anymore. Nowadays it’s more that VMs in general create a sort of IO mixer and that discerning IO patterns in VM environments is neigh impossible without insight into the VM’s IO workload in isolation.
Why is it that boot storms are no longer a concern?
It seems that a couple of things have emerged as more VDI implementations are put in place. For example, not everyone in a company boots up on Monday morning at 8am, spreading out any potential boot storms over a much longer period of time than anticipated in boot storm simulations.
Also it turns out that a lot of people never actually shut down their desktops so the need to boot is drastically reduced for these people, perhaps once/week or once/month or once/bluescreen. Although, I don’t know this for a fact, someone mentioned that VMware View has a parameter that can disable end user shutdowns and just puts the VDI instances into suspended animation (I would say sleep but that seems to be a Mac term).
Another case in point is that VMware View Planner doesn’t simulate or even measure virtual desktop boot up activity. It seems that simulating boot storms is no longer a reasonable way to help measure how effective storage systems can handle VMware View implementations.
On the other hand, I am aware of at least one other VDI benchmarking tools that make a point of simulating boot storms and other similar extreme workload activities.
So are boot storms, no longer an issue for storage systems in VDI implementations or not?
Photo Credits: Storm cloud, Duncan, Oklahoma by chascar