3 ms·

But being 95% performant compared to native speed is not worth a lot if the "only 85% performant" alternative costs millions less, since you can often spend hal

by IcePic 3y ago

But being 95% performant compared to native speed is not worth a lot if the "only 85% performant" alternative costs millions less, since you can often spend half of those millions on more or better hw and still come out with a decent profit. It's cool and great to have the best product, but if the "only" edge is that if you spend a million bucks on license, you save 10s or 100s of K in hw expenses, the business proposition is actually not that great. I get that the whole ESX ecosystem has a lot going for it, but as some of the VM loads are getting more and more dumbed down, being the one with most features and some more percent better perf is not going to win out in the end, as I see it.

nunez 3y ago
That 15% gap matters a lot for the kinds of companies that Broadcom is targeting (and many others that they are not), and they are okay with spending millions to get it. Also, you get support with those millions. That's a requirement for these kinds of companies as well. That said, I agree that there are many shops that will migrate to something like Ryzen, throw KVM on that and call it a day (or do what I did and use Graviton spot instances for anything stateful and small Graviton Lambdas for the rest). Those suckers are fast. And that's before we touch software-defined networking a la NSX and virtualized storage a la vSAN. NSX, ESX and vSAN work together to provide near-line-rate networking and networked storage performance with a really nicely packaged frontend. Yes, you can use paravirt-io and ceph and get nearly there, but it's "nearly" and setting this up has a significantly higher skill bar for sysadmins. That alone is a dealbreaker for small shops that aren't tech forward.