VSAN: is it worth it? Take licensing for example. A two-socket server is the standard for computing in the datacenter today. Standalone VSAN licensing starts at $2,500 per socket. VSAN requires at minimum of three nodes. As a result, licensing alone is just over $7000 per year.
Next, there’s the hardware. Initially, the perception is that any hardware compatible with ESXi will work with VSAN. In fact, in order to have full support from VMware with this product, all hardware items including SSD, HDD, and I/O controllers must be on the VMware VSAN HCL. This results in a price increase when purchasing hardware.VMware now strongly recommends using a fully tested configuration rather than purchasing individual components that are all on the HCL. Again, this raises the price.
Additionally, HCL compliance does not imply or guarantee a stable environment. Take for example the Dell R730 with the PERC H730 controller. That card is fully supported by VMware and yet if you look at user forums, several customers that purchased that server with that controller are having stability (PSOD) issues. An standalone hyper-converged system (i.e. Nutanix) is almost guaranteed to be more stable because of full intensive testing and control of all components (is a Windows vs. OS X comparison fair here?).
With the medium cost of entry and potential stability issues even when fully compatible, VSAN makes sense only in certain scenarios. It also make sense to diverge from the HCL slightly if needed assuming that you are aware of the risks of doing so.
In the next couple of posts, I will attempt to document my first VSAN deployment starting from scratch. I will be using some components that are on the HCL and others that are not and I will attempt to justify why I have made those decisions. Please leave comments or any concerns that you have because the more feedback the better!