[treesknees]

> KVM performance is orders of magnitude better than VMWare

Do you have any sources for this? I worked at a company developing NFV appliances, we always had much higher network throughput on VMware than we did using KVM without using some type of convoluted vswitch alternative or PCI passthrough.

VMware isn't just a hypervisor, it's an entire ecosystem of VM management and orchestration. You can tie it into AD, delegate different permissions and roles to users/groups, manage upgrades, interact with PowerShell and other APIs, it has integration into Dell and Cisco solutions, all sorts of additional features you won't find running CentOS and KVM without adding more 3rd party software on top and cobbling it together.

[gregmac]

> VMware isn't just a hypervisor, it's an entire ecosystem of VM management and orchestration. You can tie it into AD, delegate different permissions and roles to users/groups, manage upgrades, interact with PowerShell and other APIs, it has integration into Dell and Cisco solutions, all sorts of additional features you won't find running CentOS and KVM without adding more 3rd party software on top and cobbling it together.

This is it, really. For big companies this kind of stuff is important.

And "cobbling it together" is very much understating the effort involved to keep it running: eventually you'll upgrade one of the components and it will break something, because you didn't read the release notes of an upstream dependency that mentioned a breaking change that affects your particular setup.

Having the vendor (vmware) provide this as a delivered, tested, supported solution is so much easier.

Similarly, I suspect it would be significantly simpler to find IT firms and/or hire individuals with VMware knowledge than it would be to find the equivalent on KVM + Cockpit + the dozen other components you need.

Note: I'm not saying this is right or the way things should be, but simply pointing out the "enterprise" perspective. Boring technology is safe.

[foobiekr]

Same. VMware leagues better than the various horrible Openstack deployments some customers want to use.

I have in depth knowledge of kvm and the issue isn’t kvm it’s everything else.

[nimbius]

colour me unimpressed. VMWare cheerleaders haul out openstack as their touchstone example of how hard VM's are in linux but forget that VMWare pales in comparison to what Openstack is, which is an entire full-stack cloud hosting provider with accounting, DNS, tenant metering and network delegation, and support for k8s.

Start comparing VMWare against Proxmox, which is an out of the box solution anyone can use and includes every single feature of ESX and many vsphere features youd easily lose your shirt for. https://www.proxmox.com/en/

heres an independent performance test. KVM is easily faster than ESX.

https://www.spec.org/virt_sc2013/

[treesknees]

You've missed the point a bit, and for the record I prefer to deploy Proxmox/KVM over VMware most of the time, I'm not a VMware cheerleader.

There are plenty of features Proxmox doesn't have that VMware does have. I've ran into a few of them

1. No ability to pin vCPUs to physical CPUs from within Proxmox. You have to drop to bash and set affinity for each vCPU's PID by hand if you want that.

2. You can't provision a VM with more vCPUs than physical CPUs. For example if I have a host with 8 cores, the max vCPU I can allocate to a VM is 8. And yes, I did have a use case for this.

3. You can't configure networked serial ports from within Proxmox. You have to drop to bash and edit the vm configuration file by hand if you want that.

4. Lack of serial port concentrator, which means you can't really use networked serial ports reliably when migrating VMs across hosts in a Proxmox cluster. In the NFV world this can be pretty important.

5. You can't manage multiple Proxmox VM hosts from a single UI unless they're clustered, which in many cases isn't practical to do. vSphere will let you manage multiple independent hosts from a single pane of glass.

6. (at least historically) lack of RSS/multiqueue in virtio networking. vmxnet3 on VMware supports this and allows you to scale better. But I will admit it's been several years since I've had a look at this area.

Again I'm not a VMware cheerleader. I'm sure I could generate a list like this for what Proxmox has that VMware lacks. But it's incorrect to state that it includes every single feature of ESX.

[thedougd]

I've deployed the full VMware Integrated OpenStack running on ESXi, NSXT, and network attached storage; everyone was happy. Users happily deploying full stacks via API in their own Tenant with tools like Terraform. Operators thrilled to be able to patch and reboot a host without even a notification email because VMware Live Migration. The bosses without a worry because in the end it's just running on VMware.

Expensive as hell though.

[scarface74]

You don’t grok “the enterprise”. Any CTO of a large company would feel much better about having VMWare behind them and their ecosystem than your proposed solution.

“No one ever got fired for buying IBM” applies equally to big enterprise SaaS providers and companies like Oracle, Microsoft, VMWare, Salesforce, etc.

[rad_gruchalski]

I learned this lesson in an interview once. I was bullish on sap pricing and the cost of consultants. The person from top management, multi-billion chemical, said “you think we have no money”? They did, 2 years later they’ve opened a brand new building to fit a 1000 of those consultants.

[tlamponi]

Proxmox provides enterprise support staffed mostly by devs and engineers, and not some crappy outsourced service asking you the three questions from their predefined support handling sheet.

It is also deployed in many many enterprise and government settings.

[scarface74]

Again, you’re arguing about technology which is besides the point. A CTO in the enterprise is going to make the safe choice.

Also, notice when you go to VMWare’s home page you see “referenceable clients” - ie well known companies that use the software? This is “Enterprise Marketing 101”.

Besides, I can throw a stick and find someone who knows VMWare. As a (hypothetical) CTO of a non tech company, I don’t want “support”, I might want an MSP to do it for me.

[tlamponi]

This is my first reply to you, so how can I argue against your point again? Also it seems I can count myself lucky to work under a chief technology officer that actually thinks about technology and also understands actual benefits, not the ones just looking at the gardner quadrant of "who paid the most for pr". Besides that:

How vmware is a safe choice is beyond me, their proprietary so one is mostly locked in and at their bidding, their standard support reply is to install from scratch as otherwise the won't look at the setup. If you don't look at the technology POV (in addition to others) the choices probably won't be the best.

Notice how on Proxmox.com there's a testimonial page that links to enterprise and gov't customers, how's that different "Enterprise Marketing"?

Proxmox also has reseller partner all over the world: https://proxmox.com/en/partners/reseller

So we got enterprise support, world wide enterprise reseller network, enterprise features available for all, with or without support, not sure how much more enterprise vibes one can get..

vmware isn't really moving forwards since a while, the acquisition by broadcom won't help that either IMO, but sure they're currently still able to pay "bribe" some CTOs or sales people lunch, so they got that edge going for them, won't be enough to stay the #1 in the long run though.

[scarface74]

This gets back to the other boogeyman “lock in”.

The average corporation is dependent on 110 SaaS offerings.

https://www.statista.com/statistics/1233538/average-number-s...

No matter what it is a pain to migrate. Corps aren’t being bribed. They are making the safe choice.

For instance even if some unknown cloud provider is “better”, if something goes wrong, the CTO is going to be questioned. If Azure goes down and it’s already an MS shop, people aren’t going to question whether it was a bad decision.