After 114 days of change, Broadcom CEO acknowledges VMware-related “unease”

misk@sopuli.xyz to Technology@lemmy.world – 325 points –
After 114 days of change, Broadcom CEO acknowledges VMware-related “unease”
arstechnica.com
58

You are viewing a single comment

If you’re running a lab or a small shop any hypervisor can likely do the job. Anything above that VMware’s overall ecosystem is the most robust and well-supported.

At this point virtualization is a legacy technology. It’s not going to disappear tomorrow but its clock is ticking the same way the clock was ticking for mainframes thirty years ago. Plenty of mainframes still out there but nobody is implementing new. Same can be said for virtualization. It’s a limited market with significantly slowed growth over where it was a decade ago.

The move to a subscription model will let them squeeze every last dollar out of the technology while they still can.

Wait what

virtualization is a legacy technology

AWS, GCP, and Azure run on virtualization. Do you think all these cloud providers are providing everyone bare metal? This doesn’t include containerization which is a subset of virtualization. Your average shop might not run virtualization directly unless of course your team touches VirtualBox or Vagrant or qemu or (probably shouldn’t) HyperV.

Either your understanding of virtualization is very lacking or you didn’t explain your point very well. I am really curious what you meant.

Yeah, virtualization is definitely not "legacy technology" 😂

Virtualization, as a commercial product pointed at businesses, is a legacy product.

Of course large providers are utilizing virtualization, containerization and an abundance of similar technologies. However, they’re not generally using VMware to do it.

I spoke in the context of OPs question.

Maybe a better term is “commodity” technology. Rather than a thing by itself where one company is much better, now it’s everywhere, mostly good enough. It’s not going anywhere but I wouldn’t run a business on it alone

Yes commodity is a much better term here. It is a mature and fairly ubiquitous technology at this point.

I think that’s a fair point. Trying to build a new virtualization company today would have huge initial investment and a steep path to the companies that run their data centers.

(probably shouldn’t) HyperV

What makes you say “probably shouldn’t”? WSL use is widespread at this point

WSL is also shit for any kind of containerization and HyperV fucks up everything else. If you’re not doing any DevOps/SRE stuff WSL 2.0 is fine provided you don’t mix the filesystems. I have been so frustrated with their claims on release for 1.0 and 2.0 that I haven’t evaluated the recent systemd release for WSL. I provision WSL for people that don’t know why they should care and Linux VMs for people that need to work with CI tooling.

In general if you use a Microsoft tool you have to use the Microsoft ecosystem. Sometimes that’s not a huge deal, eg VS Code just adds a ton of telemetry and GitHub reads all your public code. Sometimes it’s a huge deal, eg you want to do literally anything beyond Docker Desktop defaults in the container world.

At this point virtualization is legacy technology.

Man, I'd love to believe that - and please Lemmy, prove me wrong, but virtualization, especially commercial products like VMware have one huge advantage over things like kubernetes - it's effectively plug and play and has full support available.

  1. Boot off this esxi iso
  2. Deploy this VCSA OVA
  3. Have vCenter auto config VSAN
  4. Deploy fully ha/Drs managed VMs

I would kill for a similar experience with kubes - something that I cannot for the life of me get to work in my homelab given the myriad of walkthrough in various states of accuracy.

Yeah the above is someone who can either greenfield because they work for a new shop or is deep in the kube-aid.

Most buisnesses just need your stated features. Some stand alone VMs with HA that makes them server independent and let you snapshot/back them up with ease. Slot commodity servers in a room somewhere, wander off to do more important things.

The good news is that proxmox is already there, if a bit more crunchy to deploy. It's got integrated ceph to work as a backend for HA/VSAN, built in snapshots and a separate backup appliance that sports Veeam style features. It has a simple config language that could compete with powercli if that is a current vmware use case. Looks like it even supports VDI with Intel's enterprise gpus, although that is early days.

Certainly any cloud kubernetes will be much quicker and simpler. To the extent you can use minikube, I’d call it about the same as a vm.

For large on-prem clusters, you may have a point. I did an eval and came to a similar conclusion, but that was about five years ago. Also the question was “Can we ask customers to do this or already have this for installing our product?”. That idea was a bit before its time

Please forgive a wildly uninformed question: What is it that VMware does today that isn’t covered by Docker?

Different OSes. Windows and Linux for example. No way to run a full fledged domain controller in a container. Just to name a example.

VDI environments is one place. Also Windows heavy environments (exchange, SharePoint, teams, DCs, etc) are probably better suited for VMs.

Storage, software defined networking, performance metrics, VDIs, endpoint security, virtualization on the desktop.

Not to mention, a lot of workloads aren't suited to containers. The vast majority of business software isn't containerized, and it would be wildly cost-prohibitive for me to shoehorn that square peg into the round hole of virtualization.

Uhh, docker is containers, VMware is a virtualization hypervisor?

Ultimately, in terms of what they can do, well technically you can do anything without any container or virtualization strategy, so from that angle, they are all the same.

It reallyboils down to what the humans are comfortable with, and that's where there's some divergence.

Sure, one could make technical arguments about one can be multi kennel, one has arguably somewhat stronger likelihood of isolation, and one has a bit more efficiency than the other, but it's really down to human factors and familiarity.

Developers tend to like container based approach because the "image" is transparent and usually provides nice cheap options to somewhat track history and "fork" from common points with some flexibility. VMs kind of have some of that, but practically speaking it's far more awkward.

Conversely some operators find managing container based solutions too "developery" and find comfort with virtual machines. It's also more straightforward to just carve out a vm, hand it over, and give them the keys and let them deal with it. Then you commonly have VMs at one layer, and at least some of your tenants self managing some container management layer on top of their slice of the world.

While there is some overlap, general comparison of VMware vs Docker is a bit apples and oranges.