Depends on what you define as "a toy", I suppose. "Able to handle 125 VDI instances per node without complaint" seems reasonable to me. Alternately "200+VMs across 4 nodes that run a mix of workloads ranging from Exchange to SQL to VDI to image render engines" seems not toy-like to me.
Now, to be perfectly honest, I haven't run Maxta against those controllers. Maybe Maxta would run like a dog on them too. I do know that I don't need hardware controllers to run Maxta, and it does a damned fine job using AHCI SATA 7200 rpm spindles + SSDs. It handles all the I/O I want to throw at it on a per-node basis right up to the point that I run out of RAM.
Now, it's obviously an open question how each of these will behave when we start talking about setups that run $25K+ per node just for the hardware. I can't answer that. But I do know that Maxta runs real-world production workloads just fine on some rather weedy hardware using configs that VMware officially pooh-poohs for their own offering.
It's all about what you're optimizing for. Are you optimizing for IOmeter and SPC-2 benchmarks, or for workloads that are ridiculously latency sensitive that only 0.000002% of the world actually employ? Or are you optimizing for the kinds of workloads (and disk/CPU/RAM balances) utilized by 80% of the world's businesses?
So yeah, some of this shouldn't have made it onto the HCL...but there are things that got pulled from the HCL (or never made it on) that also raise huge questions about what the merry hob VSAN is doing that it can't get workable performance out of the same hardware that is used by Nutanix (certain LSI controllers), or Maxta (AHCI).
It makes me ask uncomfortable questions. Like "is the only thing that VSAN has to offer a series of high benchmarks, and even then only when used on the absolute best of the best hardware"? How does VSAN work with the kinds of hardware normal people and mundane businesses can actually afford, or already have to hand?
If asking those questions, instead of blindly lapping up marketing tripe and praising a solution that is twice as expensive as others competing offerings is "not understanding storage basis" I'm remarkably cool with that.
"How many IOPS do you need" is just as important a question as "how many IOPS can this solution deliver." And what's really of interest is the lovely question "why can X deliver Y IOPS on Z hardware, but W cannot?"
And, quite frankly, I don't care who gets upset when I ask those questions.