I get why enterprise likes VMWare but KVM isn’t harder to deal with. I’ve always worked at smaller companies so this isn’t an expert opinion. But I’ve always felt like at the infrastructure level, it ends up being cheaper to hire experts and run the open source solution (assuming it’s mature and at feature parity) than pay licensing and support fees.
An expert in one thing will usually add to your company in other ways too. Talent > “solutions” in the long run.
I learned years ago that buying something from a big company instead of using a free open source solution is about aoutsourcing responsibility. Its about being able to sue a company about damages instead of hiring reliable personell to run and write fixes for foss software. Also insurance is much easier.
Not that I follow that advice, my company is still 99% foss software.
Exactly. All enterprise software pricing is about liability and how much blame you can put on the software vendor when everything goes wrong.
Of course, your company’s employees are still going to be the ones picking up the pieces but you get to tell your investors “actually it was VMWare’s fault”. And you can’t do that with OSS, even if you’re buying support.
Let me tell you about a large bank and two data centers operated using VMware and the type of talent the bank is able to hire and retain. A move away from VMware is a 5-year project involving hiring, retraining, design mistakes, budget overruns, and a lot of grey hair. The year was 2012. 7 years later, one DC converted to OpenStack, the project is shelved and the majority of th OpenStack DC gets converted back to VMware due to “OpenStack disaster.”
This one gave me the imposter syndrome rollercoaster again. I’m inept, you can’t convince me I’m not, but I’m not that inept that I can’t see multiple glaring flaws with different explanations at play here. That passes for IT professionals in pivotal roles. But I’d be too embarrassed to continue existing if I were anyone involved in that rollout “plan”.
Obviously there’s some sort of spectrum of knowledge and experience, but I cant reliably discern what it’s supposed to even look like anymore.
You’ve got an audience if you feel like elaborating. Sounds like a fun story haha.
The fundamental problem I have identified over the years I worked adjacent to this project is this. Most folks above the manager position are not technical. They’re typically some sort of BA. These kinds of folks do not easily comprehend the technical merits of different solutions. All sorts of errors stem from that. Errors in estimating risk, errors in estimating difficulty, but crucially errors in telling reality from fantasy, or truth from lies. Under this framework, the ability of the organization to hire technical people who know what they’re doing is more or less based on luck. This particular org struck luck with some hires and didn’t with most. So now we have a group of people who will build this thing, with only a few qualified people among the unqualified. Alright. A difficult design decision has to be made. There are two proposals. One from a qualified person. Another from an unqualified one. They’re both presented to a director or a VP for a final decision. The qualified person presents their design, pros, cons, etc. The unqualified one does the same, except they have an ace up their sleeve - confident lies. So they sprinkle those all around their design - everything is amazing, few cons if any, unicorns shitting rainbows and the lot. The decision maker cannot discern the lies from the truth. The unicorn design feels irresistible. It’s chosen. Its designer is promoted before its ill effects are ever realized. Now the competent folk don’t even get to present alternatives to the VP level. Eventually they’re tired of this shit and move to a place that is less corrupted.
This wasn’t confined to one project and a single set of people. It’s a general problem that transcends orgs and companies.
I work in hosting. We mostly use Proxmox for our Hypervisors which is already a step up from “bare” KVM in regards to convenience/ease of use (especially for High availability scenarios and the like)
We also run VMWare and while I don’t love the “locked down you gotta do it the VMWare way” nature it’s often so much easier and the HA is mich more convenient. Also it has proper functionality for custom resourcing/access/billing.
Say what? Going back to only KVM in modern DCs is some crazy talk. If your org is small enough that KVM is even remotely an option, then I’d recommend running a cost/benefit analysis on whether hosting a small server farm on prem is even worthwhile.
But when you’re managing hundreds of servers with dozens of various purposes, FOSS solutions aren’t always tenable. And not using a mature, feature complete virtualization platform is just straight up masochistic, not to mention potentially dangerous from a security standpoint.
I agree that talent > solutions, but if you want to retain that talent, you have to make their lives not miserable at work, which means sometimes having to purchase solutions to make their lives easier.
Small and medium (and even large) companies investing in talent instead of commercial solutions is the solution to improving FOSS. I know it has downsides, as you stated, but there are significant upsides. FOSS is cheaper than a custom solution, and the company only has to pay for the modifications it wants to see. The whole community then benefits from their hard work adding features and maintaining the software.
I’m not saying that it’s the BEST idea for every company. All I’m saying is not to discount FOSS out of hand for these companies. There are significant advantages for companies that should be weighed against the cons. This kind of advocacy is also important in furthering the FOSS model.
I get why enterprise likes VMWare but KVM isn’t harder to deal with. I’ve always worked at smaller companies so this isn’t an expert opinion. But I’ve always felt like at the infrastructure level, it ends up being cheaper to hire experts and run the open source solution (assuming it’s mature and at feature parity) than pay licensing and support fees.
An expert in one thing will usually add to your company in other ways too. Talent > “solutions” in the long run.
I learned years ago that buying something from a big company instead of using a free open source solution is about aoutsourcing responsibility. Its about being able to sue a company about damages instead of hiring reliable personell to run and write fixes for foss software. Also insurance is much easier.
Not that I follow that advice, my company is still 99% foss software.
Exactly. All enterprise software pricing is about liability and how much blame you can put on the software vendor when everything goes wrong.
Of course, your company’s employees are still going to be the ones picking up the pieces but you get to tell your investors “actually it was VMWare’s fault”. And you can’t do that with OSS, even if you’re buying support.
Let me tell you about a large bank and two data centers operated using VMware and the type of talent the bank is able to hire and retain. A move away from VMware is a 5-year project involving hiring, retraining, design mistakes, budget overruns, and a lot of grey hair. The year was 2012. 7 years later, one DC converted to OpenStack, the project is shelved and the majority of th OpenStack DC gets converted back to VMware due to “OpenStack disaster.”
This one gave me the imposter syndrome rollercoaster again. I’m inept, you can’t convince me I’m not, but I’m not that inept that I can’t see multiple glaring flaws with different explanations at play here. That passes for IT professionals in pivotal roles. But I’d be too embarrassed to continue existing if I were anyone involved in that rollout “plan”.
Obviously there’s some sort of spectrum of knowledge and experience, but I cant reliably discern what it’s supposed to even look like anymore.
You’ve got an audience if you feel like elaborating. Sounds like a fun story haha.
The fundamental problem I have identified over the years I worked adjacent to this project is this. Most folks above the manager position are not technical. They’re typically some sort of BA. These kinds of folks do not easily comprehend the technical merits of different solutions. All sorts of errors stem from that. Errors in estimating risk, errors in estimating difficulty, but crucially errors in telling reality from fantasy, or truth from lies. Under this framework, the ability of the organization to hire technical people who know what they’re doing is more or less based on luck. This particular org struck luck with some hires and didn’t with most. So now we have a group of people who will build this thing, with only a few qualified people among the unqualified. Alright. A difficult design decision has to be made. There are two proposals. One from a qualified person. Another from an unqualified one. They’re both presented to a director or a VP for a final decision. The qualified person presents their design, pros, cons, etc. The unqualified one does the same, except they have an ace up their sleeve - confident lies. So they sprinkle those all around their design - everything is amazing, few cons if any, unicorns shitting rainbows and the lot. The decision maker cannot discern the lies from the truth. The unicorn design feels irresistible. It’s chosen. Its designer is promoted before its ill effects are ever realized. Now the competent folk don’t even get to present alternatives to the VP level. Eventually they’re tired of this shit and move to a place that is less corrupted.
This wasn’t confined to one project and a single set of people. It’s a general problem that transcends orgs and companies.
RHEV > OpenStack
Don’t hate the game just because of one frail player.
But isn’t RHEV dead?
I work in hosting. We mostly use Proxmox for our Hypervisors which is already a step up from “bare” KVM in regards to convenience/ease of use (especially for High availability scenarios and the like) We also run VMWare and while I don’t love the “locked down you gotta do it the VMWare way” nature it’s often so much easier and the HA is mich more convenient. Also it has proper functionality for custom resourcing/access/billing.
Say what? Going back to only KVM in modern DCs is some crazy talk. If your org is small enough that KVM is even remotely an option, then I’d recommend running a cost/benefit analysis on whether hosting a small server farm on prem is even worthwhile.
But when you’re managing hundreds of servers with dozens of various purposes, FOSS solutions aren’t always tenable. And not using a mature, feature complete virtualization platform is just straight up masochistic, not to mention potentially dangerous from a security standpoint.
I agree that talent > solutions, but if you want to retain that talent, you have to make their lives not miserable at work, which means sometimes having to purchase solutions to make their lives easier.
Small and medium (and even large) companies investing in talent instead of commercial solutions is the solution to improving FOSS. I know it has downsides, as you stated, but there are significant upsides. FOSS is cheaper than a custom solution, and the company only has to pay for the modifications it wants to see. The whole community then benefits from their hard work adding features and maintaining the software.
I’m not saying that it’s the BEST idea for every company. All I’m saying is not to discount FOSS out of hand for these companies. There are significant advantages for companies that should be weighed against the cons. This kind of advocacy is also important in furthering the FOSS model.