Broadcom loses another big VMware customer

When it was time for our contract to renew a few months back, our licensing and support quote was 3x the previous year and Broadcom would not budge, even a little. They said take it or leave it. Well, we left it.

There is a big internal effort now to get us off of VMware and onto Kubernetes and OpenShift. Our whole fleet of VMware is still running but we're on borrowed time as we're on our own if any major technical issue comes up.

I suspect Broadcom's plan is to figure out just how much they can bear and cut a deal along those lines.

That would totally work -- we're neck deep in vmware, and there's no easy way out -- except that there's a kind of network effect at play.

Support is huge for us. In the beginning, we developed expertise in vmware (particularly in regard to our products), and guided our customers to it as well. Customers were happy to follow our recommended/supported option, and we were happy supporting it. (The support ain't cheap!)

But our customers, by-and-large, are smaller and somewhat more nimble, meaning many will be unable or unwilling to do a deal with broadcom, and will insist on switching away. I suspect new customers will lead the way... there's no way we're going to walk away from a deal because the customer refuses to pay for vmware, and we probably can't afford to eat that cost for them. So we're going to start gaining expertise and supporting other vm platforms. I think we'll follow our original pattern: settle on one, gain expertise, start using it in support, then test and dev, and steer our customers to it. At some point we'll realize we know how to get rid of the rest of our internal use of vmware, AND realize we can sell remaining customers support to switch away from vmware themselves. Once that happens, I think it becomes a ball rolling down hill, and things will accelerate quickly.

There are a couple reasons this isn't like Oracle, BTW. For one, with Oracle we're 10,000 ft below the surface with Oracle-provided air tanks. Second (I guess this is the main thing), Oracle is our problem, not our customers'.

It's just further re-emphasizing that it doesn't matter how good your vendor is now, they will probably eventually get acquired by a company attempting to squeeze every last penny out of you. And their cost-benefit calculation is no longer based on how much unique value the software is providing, but it's instead about how much of a PITA it is to migrate off said software.

Even in the early 2015's kubernetes was mature enough to run production workloads. Any issues that did come up were with our own application code, never the orchestrator.

I never found myself thinking that it would be nice to have a close sourced expensive option to reach for.

What area does VMWare excel so much to justify this pricing power?

No real surprise there. As long as the bill for infrastructure remains sane, nobody is going to put in the effort to change out important parts of your arrangement, instead sticking with a "Well, we know this works, and we know how to deal with it..." approach.

If they'd raised the bill somewhat - 50%, 100%... people probably would have stuck with it. But to jack it an order of magnitude, well, now it's worth putting the engineers on the project to find a cheaper solution (that may very well be better - virtio vs what VMWare is using... I certainly prefer virtio for most of my storage and networking needs).

> The tech team also warned management that the quality of VMware's support services and innovation were falling.

I mean, the writing was on the wall, you don't buy out a product and jack the prices 10x if you plan to actually support it. It's pure "value extraction" at that point. Sad, really, because VMWare has a lot of nice features behind it and has been a well thought out bit of virtualization software throughout the years.

I had the displeasure of having to update a VMWare install on a laptop recently (VMWare Player had been perfectly fine, which was discontinued, Workstation is now free for personal use, but you have to register with your full physical address to download it, and I just want to run the VM I use to talk to my car, please...). I can't say I'll be considering them for anything going forward.

They are the second shittiest company I’ve ever dealt with — only recently outdone by Intel/Altera.

Broadcom bought VMWare knowing they'd trap most into existing contracts as whales, and knew most would bear the cost increase regardless.

Few casualties, so what - look at them profits.

VMware's biggest problems remain its lack of product cohesion and complexity relative to its cost, and while their recent conference suggested they're working to address these concerns, I worry it's about half a decade too late for it to be of value. Aria should have been a single product suite and appliance when they pivoted to public cloud, and vRealize Automate shouldn't have been so overly complex as to require professional services for deployment and maintenance. NSX seems powerful on the surface, but really only seems to thrive when the network team either relies upon it directly, or gives you a widely trunked subnet to work with and carve out. vSphere and ESXi are excellent stalwarts, but they're not as usable/automation-friendly as they should be in the modern era of IaC. Pre-Broadcom, VMware's strength was its ubiquity and core feature set for VMs relative to its pricing; post-Broadcom, not so much.

As for the alternatives, I'm not really sold on any of them from commercial vendors.

* Nutanix has never turned a yearly profit, doesn't support commodity hardware very well (especially if you want the full feature set), and is already expensive on its surface. The pivot towards cloud-hosted services in lieu of on-prem services suggests it's trying to emulate larger and more successful competitors, rather than focusing on its core business strategy (HCI). It's a shame, because if they went the VMware route (software only on commodity hardware), I think they'd be more successful.

* Virtuozzo was the best commercial competitor to VMware on paper, but I never got to do a proper PoC. They're more focused on consumption-based services and XaaS, which is a plus, but I'm cautious endorsing them further as I have no direct experience with them - though I did suggest a PoC at the time, to explore their IaaS and PaaS suites.

* OpenShift, while a very neat tool (VMs managed the same way as Kubernetes, huzzah!), doesn't really fit modern Enterprise needs either. I applaud any tooling that tries to make infrastructure scalable like containers are, but Enterprise workloads remain "VM first" for the most part, which OpenShift isn't really aiming for. If you're already working mostly with containers, I'd strongly suggest evaluating OpenShift, but most companies aren't, and Red Hat/IBM is very focused on selling VM-heavy customers on this container transition despite most of our workloads not supporting containers at all, and those that are available as containers often have strong warnings against orchestration support with K8s/OpenShift.

* Microsoft's own stack is kind of a known quantity that I'd really only recommend if you're already a huge Microsoft customer, or rely heavily on MSPs/outsourced labor to support it. If you're not already on Microsoft, there's no reason to switch to it.

* Proxmox was considered, but didn't make it into my final proposal due to prior evaluations not finding it Enterprise-suitable.

* Red Hat Cockpit + KVM is surprisingly powerful! It'd be my default recommendation for SMBs, as RHELS pricing is fairly comparable to what vSphere licenses used to be, and once your engineers are onboarded into the world of Linux, they can convince you to migrate distros to something cheaper without significant disruption.

When looking through free/Open Source projects, I initially narrowed it down to two:

* Apache CloudStack was my 1st choice in my research, and the PoC was great. The initial buildout can be annoying, as it's typical Apache faire (in my experience anyway) of multi-page-manuals for a "quick start", but once the initial manager is up and running, the rest moves pretty quickly.

* OpenStack was considered, but its complexity and composition of dozens of individual projects made it untenable to support in our environment.

* OpenNebula was not considered at the time, but I've been looking into it on my downtime and considering my own PoC, since I dislike Proxmox for my homelab.

My instincts tell me that we're a few years out from a "great reshoring" of workloads into more appropriate placements: companies reluctant to move to public cloud will do so for customer-billable workloads because they can have an easier time judging/fixing margins on it and scaling to demand, while Enterprise and stable workloads will likely move back on-prem where sovereignty and security are paramount and cost savings can be achieved through longer hardware lifespans. In that context, nobody does a particularly good job of creating a "Universal Cloud" abstraction layer to free engineers from juggling dozens of APIs, CLIs, codebases, and pipelines for each specific vendor or workload; not even Kubernetes does this well, and it's arguably the best presently-available technology suited to cross-cloud management and orchestration. Broadcom jumped the gun on price increases, because given another year or two of product improvement and shifting landscapes of both technology and geopolitics, they'd essentially have a captive audience to squeeze for higher margins; instead, we have ample time to consider alternatives that are quite comparable to VMware in capabilities, and often substantially cheaper to boot.