VMware's Stumbling Cloud Adventure
chrisdodds.net
chrisdodds.net
@OP HN guidelines say "please use the original title, unless it is misleading or linkbait" https://news.ycombinator.com/newsguidelines.html
> Unfortunately, the world that they built is now eating them. Hypervisors became commodity, where "good enough" is an acceptable target. Hyper-V, Xen, KVM: they all became good enough for ecosystems to be built around them
In reality VMware (and VirtualBox) were leapfrogged by platform native hypervisors quality-wise, at least on Linux. Which is inevitable since there isn't really any way to ship robust proprietary binary-only kernel drivers of this complexity there.
(ESXi might be another story, but it's also not just a hypervisor, it's a whole stack of enterprisey proprietary infrastructure that you commit to instead of open systems, and hopping on that train is more a cultural/strategy/mindset choice than something based on technical hypervisor-vs-hypervisor benchmark)
I may be wrong here, but I thought ESXi was the lightweight hypervisor-only product (which at the time was offered for free!), without all the enterprisey features or requirements.
Basically Microsoft Windows Hyper-V without the need to boot an OS of its own.
VMWare started doing this circa 2010-11.
Eventually that project turned into the first version of Cloud Foundry.
Edit: forgot my usual disclaimer that I work for Pivotal etc etc
Disclaimer: Currently working with a customer deploying to Cloud Foundry.
I usually refer to Cloud Foundry as a PaaS. It sits on top of an IaaS like AWS, OpenStack, Azure or vSphere, managed by BOSH.
http://arstechnica.com/information-technology/2016/01/vmware...
So they've lost the cloud... They're firing the desktop guys. Where do they plan to go?
They couldn't go from a lucrative private cloud revenue stream and then somehow convert it to a public cloud and expect to make the same amount of money. During that time that was mentioned, from 2006 until probably 2010/2011, AWS was great for small companies, but there were still issues with security, availability, reliability, etc, so VMware could get away with selling the private cloud.
I mentioned this is another post that got a lot of vehement comments, but virtualization itself is dying. Not dead, but dying. Many large companies with their own datacenters are running off of bare metal, and this will be the trend going forward. Yes, virtualization is still in many private datacenters, but bare metal is how newer companies are going, and in 10 years, I bet virtualization will be gone from all but the slowest of private datacenters. It's a level of overheard now that has been completely circumvented as unnecessary for the most part, and it will fall into disfavor as more and more people learn techniques on how to avoid it altogether.
This doesn't mean that VMware will be gone in 10 years. Even WinZip to this day has something like $50-75M/year in annual revenue. But it will be relegated as a legacy app.
I know of some that are deploying some things back to "bare metal" if that's what you want to call containers, but even in those instances, it's a very, VERY small subset of the larger environment.
That said it is usually deployed beside a cloud like thing, and only used for select workloads.
Do you have any thoroughly researched numbers to back up this claim because I really don't see the evidence and I look at quite a few companies every year?
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.
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.
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.
Is it? To me Mesos is a system for job management across a heterogenous pool of compute resources.
"What does that mean?"
"I spend all day thinking about what to name things."
reminds me about Sun. There i learned that if you don't cannibalize your main source of revenue yourself, somebody else will do it anyway.
>They couldn't go from a lucrative private cloud revenue stream and then somehow convert it to a public cloud and expect to make the same amount of money.
it is the way of technology - from higher margin low volumes things grow/scale into lower margin higher volumes (with total money on the table only growing). To play that transition successfully a fat slow cow has to start moving faster and to become more lean.
>virtualization itself is dying
the original way - full VM like the VMWare Workstation - has become just one among several. Containers are here to stay for enterprises. Many enterprises aren't going to bother how you run the service which runs their containers as long as it runs and scales.
And on top that bare metal is ESXi. Not all companies run cloud services or large web-apps. Most run Exchange, database servers, SharePoint, CRM or ERP systems. For these companies virtualization is a way of saving rack space, while getting the benefits of redundancy, failover and the ability to quickly spin up a new server.
AWS is barely in the same business as vmware, even though Amazon would like to convince us otherwise. AWS is simply not an option for a ton of companies that don't want their data leaving their premises.
Vmware could have a nice niche business just making ESXi, but I doubt their management want to be a niche business.
How many of those companies actually were the kind of company that otherwise would have bought VMWare products? Those that have the expertise to properly move off virtualization right now probably didn't bother with VMWare.
What does "bare metal" actually mean? I don't think an embedded hardware guy has the same definition of bare metal as someone doing things in the cloud.