[pfarnsworth]

I find it fascinating that people react so violently to the notion that virtualization is dying. And being ex-VMware myself from the pre-IPO days, there's no need to sell me on the virtues of virtualization.

The point is, when you're talking about building up large datacenters from scratch, virtualization is an unnecessary expense. My ex-VMware coworkers have spread through the Valley, and when I chat with them, they are saying the same things.

Apple Siri is 100% on bare metal and Mesos. My friends at Uber also say the same thing, that they don't use virtualization at all, and the ex-VP of engineering of VMware is there now. My friends at HDP who see a lot more customers and installations that I do say that they haven't seen any of their customers using virtualization at all, they see bare metal and mesos/yarn. Someone on the previous thread mentioned Spotify is the same way.

If you're building up a datacenter from scratch, the trend is to now use bare metal and avoid virtualization altogether. Sure, if you're currently a datacenter with virtualization, you're not going to rip it out. As I mentioned above, it's not dead, and it will still be around 10 years from. But it will be relegated to a legacy app.

[mugsie]

Mesos is a form of virtualisation - just a different type.

I am not surprised HortenWorks customers are on containers / bare metal - the over head for their workloads that would be caused by a traditional hypervisor would be costly.

For single tenant DCs - there might be less virtualisation in coming years, but Mesos / Docker have not yet implemented the same layer of isolation that KVM / Xen / ESXi have, so we will still see (traditional) virtualisation in shared environments.

Also - Amazon / Azure / Rackspace are quite definitely not going to have it relegated as legacy.

[pfarnsworth]

Those aren't private datacenters. Amazon/Azure/Rackspace/Google Cloud are public datacenters providing services for customers, so they will obviously be using virtualization. Personally I think public cloud is going to kill private cloud altogether, but that's an entirely different discussion. But in the specific case of private datacenters, I believe, virtualization is dead, and many of my colleagues share the same opinion. Just because it's not common around the world doesn't mean that the change in trend and thoughts hasn't hit an inflection point. Virtualization has done an amazing job, and there's a reason why it has become so popular. But just like all other technologies, I think its going to be passed for something simpler that scales better.

[bones6]

I think you might be more accurate to say 'dead among highly innovative startups and tech giants'. You are probably totally correct long term on the dying, but it's going to be a few decades before Fortune 1000 and especially healthcare conglomerates move everything away from private cloud virtualization. Inflection points aside, that massive change takes a long time to be realized.

Your original comment makes it seem like it's all passé and old hat. It's going to be part of business in general for a long time and someone is going to have to work on it and develop it until it's gone.

[pfarnsworth]

From experience, trends change very quickly. Sun went from one of the 4 Horsemen of the Internet to dead in 10 years.

[jacques_chester]

> Mesos is a form of virtualisation

Is it? To me Mesos is a system for job management across a heterogenous pool of compute resources.

[abiox]

to me, your descriptor could perhaps refer to a kind of application virtualization.

[jacques_chester]

Someone once asked me what I do. "I'm a software engineer".

"What does that mean?"

"I spend all day thinking about what to name things."