What are you replacing vSphere with? Or: Broadcom gets absurd on pricing.

Jump to latest Follow Reply
Arkannis

Arkannis

Ars Tribunus Militum
1,977
Subscriptor
  • #1
    At work, we did the back of the napkin math and realized that our licensing and support costs are about to increase from about 14k a year to close to 90k a year with the new per core pricing that Broadcom is implementing.

    Bottom line, we aren’t able to afford it. Anyone in a similar situation, and if so, what product are you moving your virtualization stack to?
     
    Paladin

    Paladin

    Ars Legatus Legionis
    32,552
    Subscriptor
  • #2
    The big ones I have looked at are:

    Virtuozzo (or similar Openstack 'flavor')
    Nutanix
    Hyper-V
    XCP-NG/Xen Orchestra
    Suse Harvester
    Proxmox
    Azure HCI Stack

    Maybe a couple of others I am forgetting.

    Obviously VMWare wants you to move to a cloud IaaS provider that will pay Broadcom the big bucks and charge you by the pound for everything you use. So there are those too (VMWare on Google/AWS/Azure).
     
    • Like
    Reactions: WingMan
    O

    oikjn

    Ars Scholae Palatinae
    969
    Subscriptor++
  • #4
    @WingMan my understanding with Starwind is that they have a VSAN product and an HCI appliance, but I don't think they have a VM platform itself, it runs VMWare/Hyper-V/KVM VMs. Back in the day, I played around with them before using the Lefthand Networks VSA, but again, that was all just a platform for shared storage which doesn't really replace vCenter/Esxi itself.
     
    Paladin

    Paladin

    Ars Legatus Legionis
    32,552
    Subscriptor
  • #7
    From their page: "A choice of the supported hypervisors (VMware vSphere, Microsoft Hyper-V, or our own flavor of KVM), in-house developed storage virtualization software called StarWind Virtual SAN, pushing data over the proven iSCSI or cutting-edge NVMe-oF, with the help of RDMA technology (if available), an intuitive web management, an AI-powered sophisticated telemetry, and “call home” system."

    So they will run Hyper-V or VMware vSphere or their own customized KVM. That seems like it should at least start to be an option to replace VMware, if you are ok with migrating to KVM.

    I wouldn't call it the best option. I would assume they are more focused on the storage aspect than the rich hypervisor experience aspect. But it technically is there...
     
    • Like
    Reactions: r0twhylr and WingMan
    R

    r0twhylr

    Ars Tribunus Militum
    2,131
    Subscriptor++
  • #10
    TrueNAS Scale could fit the bill as well. Similar to Starwinds, it is primarily clustered storage but allows you to run KVM and containers on top.

    The big questions I would ask are:
    • Do you just need a hypervisor with vmotion and HA? Do you need other stuff like distributed switches, NSX, etc.
    • What is your backup app & DR strategy? How will changing hypervisors change that? How much do you care?
    • How many of your VMs are some flavor of Microsoft server? What MS licensing to you have?
    • Are you considering hardware replacement at the same time?

    I see a lot of shops who already own Windows Server Datacenter licenses, and Veeam or similar for backup. Generally changing to Hyper-V or Azure Stack HCI is the least problematic for them. I've seen some traction with Nutanix.
     
    Paladin

    Paladin

    Ars Legatus Legionis
    32,552
    Subscriptor
  • #13
    r0twhylr said:
    I see a lot of shops who already own Windows Server Datacenter licenses, and Veeam or similar for backup. Generally changing to Hyper-V or Azure Stack HCI is the least problematic for them. I've seen some traction with Nutanix.
    Click to expand...
    That Windows Server situation is such an interesting thing. I think if Microsoft had really put some work into a better version for SCVMM and just put Windows Admin Center in as the default web management interface on any Hyper-V/Core install so you would have an out of the box management interface to rival ESXi and vCenter, they could have made a serious stomping of VMWare but I think they just really didn't want to and must have seen early on that everything-as-a-service subscription model was the only viable future for propping up the next 10-15 years of investor share growth (which is the only thing most companies really care about these days).

    So many corporate decisions seem to be made more on the question of 'what will get me a few years of huge bonuses before I sell out and kick off to some other company to pillage for share holder benefit?' rather than 'what actually helps my customers and this company reach long term success?' Sad.
     
    • Like
    Reactions: EnglishBob, IncrHulk, Lurch and 2 others
    R

    r0twhylr

    Ars Tribunus Militum
    2,131
    Subscriptor++
  • #14
    Paladin said:
    That Windows Server situation is such an interesting thing. I think if Microsoft had really put some work into a better version for SCVMM and just put Windows Admin Center in as the default web management interface on any Hyper-V/Core install so you would have an out of the box management interface to rival ESXi and vCenter, they could have made a serious stomping of VMWare but I think they just really didn't want to and must have seen early on that everything-as-a-service subscription model was the only viable future for propping up the next 10-15 years of investor share growth (which is the only thing most companies really care about these days).

    So many corporate decisions seem to be made more on the question of 'what will get me a few years of huge bonuses before I sell out and kick off to some other company to pillage for share holder benefit?' rather than 'what actually helps my customers and this company reach long term success?' Sad.
    Click to expand...
    I've never personally worked with Hyper-V or gotten any certs on it. By reputation it's not quite as good as ESXi/vCenter at a lot of things, but it's generally good enough. Back about 12ish years ago (remember the whole VMware VRAM licensing debacle in version 5.x?) MS could have made huge inroads into VMware's market share, but honestly it kind of felt like they kind of pulled their punches competing, which is not a very MS thing to do. I think they realized that since almost every VMware sale created demand for Windows Server Data Center licenses, they could let VMware do the hard business of selling, and MS would still make bank on the licensing. That, and they were leaning hard into Azure instead of on-prem virtualization at the time.

    Tomorrow I have a call with the CTO of a business we sold a VMware ELA to last year. Fun.
     
    • Like
    Reactions: Paladin
    D

    DrWebster

    Ars Praefectus
    3,770
    Subscriptor++
  • #15
    Paladin said:
    So many corporate decisions seem to be made more on the question of 'what will get me a few years of huge bonuses before I sell out and kick off to some other company to pillage for share holder benefit?' rather than 'what actually helps my customers and this company reach long term success?' Sad.
    Click to expand...
    This is honestly why I'm making plans to retire early, or at least consider a different career later on. IT admins are always left holding the bag after corporate shenanigans, and it's only going to get worse.
     
    • Like
    Reactions: EnglishBob, WingMan, Paladin and 2 others
    Whittey

    Whittey

    Ars Tribunus Militum
    1,849
  • #17
    I don't see much of an out for us right now. We have significant investments in stretched clusters using SAN, and none of my servers have local disks, so any kind of converged environment is a no-go.

    As for saving cash, we're going to end up dropping down some licenses. We historically used vCloud Suite Standard, as we use vsphere, esxi, and vrops (aria operations). Given that vSphere Foundation (VVF) costs 3x as much as vSphere Standard (VVS), and the only addition we'd use is AriaOps, it looks like I'll be going VVS and just finding a replacement for AriaOps. Saving money left and right! :eng101: Finding a tool that works well for monitoring at the cluster level has been interesting. Lots of pretty bad products out there. Sexigraf looks neat so we'll see...

    I do have a hope that Nutanix will one day support external storage and something analagous to a datastore inside AHV. I could go with that.
     
    K

    kperrier

    Ars Legatus Legionis
    20,050
    Subscriptor++
  • #18
    r0twhylr said:
    MS could have made huge inroads into VMware's market share, but honestly it kind of felt like they kind of pulled their punches competing, which is not a very MS thing to do.
    Click to expand...
    If you think about it from an anti-trust viewpoint MS not investing that much into the platform becomes reasonable. Now, with the cloud/Azure/etc. beefing up their investment into Hyper-V and its management, bonues points if it looks like Azure scaled down to what you have, makes a lot of business sense, IMO.
     
    • Like
    Reactions: r0twhylr
    W

    waqar

    Ars Praefectus
    4,216
    Subscriptor
  • #19
    hearing good things about openNebula. Haven't broken the ice on it yet myself.
    For serving up vm's proxmox does do it.
    Its all going containerisation any how medium term. That tech is mature, running virtual machines, there's not too many if any use cases where a containerised app running on kubernetes or some such is going to provide substantially great automated resilience, and scalability
    Been thinking they are the new cpm for a while now, thinking about it, more like the new AS400 or tandem.
     
    Last edited:
    A

    acheron76

    Smack-Fu Master, in training
    1
  • #26
    Whittey said:
    I don't see much of an out for us right now. We have significant investments in stretched clusters using SAN, and none of my servers have local disks, so any kind of converged environment is a no-go.
    Click to expand...

    Whittey said:
    I do have a hope that Nutanix will one day support external storage and something analagous to a datastore inside AHV. I could go with that.
    Click to expand...
    Same here. I hope that, now that broadcom does its broadcom thing, more alternatives will emerge. Before the broadcom thing, competing with vmware in their home territory probably seemed to be pointless to a lot of others I guess, but now it could be a opportunity.

    I'm pretty sure that in 12 or 24 months from now, there'll be more alternatives to vmware than we see now. Question is, can you survive with new broadcom fees until then.

    I'd hate to switch to an alternative that only fits 70%, just to see a new, better fitting supplier showing up right after we finish the transition.
     
    K

    koala

    Ars Tribunus Angusticlavius
    7,579
  • #30
    We need some low-scale VM capability on cheap hardware (e.g. a couple of beefy Hetzner servers). I am doing a proof of concept using EL9 using cockpit-machines. Apparently it's the Red Hat-suggested alternative for this kind of scenario.

    It's nice- Hetzner can do an automated installation of EL9, then I have a playbook that sets up cockpit-machines. Then I can create Ubuntu 22.04 hosts using cloud images decently, although it has some rough edges (DNS is not working perfectly OOB).

    I am looking now to see if I can provision VMs from Ansible too- I'm finding that less documented, unfortunately.

    Doing this before going for Proxmox, which I'm more familiar with, because installing Proxmox in Hetzner has more friction (e.g. you need KVM installation or ugly qemu/vnc hacks) and I want something turnkey-er, but likely I'll go with Proxmox if cockpit-machines gives me more trouble.

    (Interested in other approaches. This is a very small org, which unfortunately needs disposable Linux boxes with 32-64Gb of RAM and 400Gb of HDD- in this space, exploring Hetzner makes plenty of sense even though it requires more engineering time to set up.)
     
    B

    Burn24

    Smack-Fu Master, in training
    53
  • #32
    koala said:
    We need some low-scale VM capability on cheap hardware (e.g. a couple of beefy Hetzner servers). I am doing a proof of concept using EL9 using cockpit-machines. Apparently it's the Red Hat-suggested alternative for this kind of scenario.
    Click to expand...
    I used the 'old' setup RHEL/CentOS used for this kind of thing with libvirt/virt-manager and it worked really well. It wasn't really scalable, but if you're doing a small volume of changes to the deploy VMs it should work pretty easy with some simple scripts and a crib sheet to do common things. 10 years ago this was a common workflow and part of cutting-edge managed virtualisation stuff, I can't imagine it has atrophied too much.
     
    K

    koala

    Ars Tribunus Angusticlavius
    7,579
  • #33
    Burn24 said:
    I used the 'old' setup RHEL/CentOS used for this kind of thing with libvirt/virt-manager and it worked really well. It wasn't really scalable, but if you're doing a small volume of changes to the deploy VMs it should work pretty easy with some simple scripts and a crib sheet to do common things. 10 years ago this was a common workflow and part of cutting-edge managed virtualisation stuff, I can't imagine it has atrophied too much.
    Click to expand...
    I'd really like a web UI instead of virt-manager. Basically the web console would be for basic stuff (sometimes people will need to access a graphical application, so viewing the physical display on a web UI would be easy. Also perhaps stop/starts/removals. And taking a look at loads...)

    Management of everything will likely be done using Ansible.
     
    waubers

    waubers

    Ars Tribunus Angusticlavius
    7,227
    Subscriptor++
  • #34
    While I no longer work there, I'm shocked there isn't more traction for Nutanix in this discussion. I saw first hand how they were able to compete and frequently beat Storage Vendor + vHosts + VMware pricing, and the product is very mature at this point. Their support is, or at least was, really good, and the supermicro boxes that they themselves sell are fine for most orgs.

    Hell, you have Cisco running aroung getting paid on Nutanix deals, so if you want a really good price, get Cisco to price it for you, the incentives being given to the Cisco reps to sell Nutanix is bonkers. At the same time, the rick of Cisco buying them might be enough for me to not want anything to do with Nutanix :)
     
    O

    oikjn

    Ars Scholae Palatinae
    969
    Subscriptor++
  • #36
    that and for an SMB, you really had to go all in and that only made sense when you were replacing all of your hardware at the same time. They really didn't have a real legitimate "small" appliance since you really needed multiple appliances to startup a system and by the time that was all done, it was a 70k+ project.
     
    • Like
    Reactions: SandyTech
    D

    DrWebster

    Ars Praefectus
    3,770
    Subscriptor++
  • #37
    SandyTech said:
    The problem a lot of my clients run into with Nutanix (and HCI in general to be fair) is that storage growth often outpaces compute growth which, at least last time we brought them in, Nutanix did not have a satisfactory answer to. Especially on the timelines SMBs tend to keep infrastructure around for.
    Click to expand...
    This is why I have minimal interest in Nutanix. Need more storage? Gotta buy another box. Need more compute? Gotta buy another box. Want to refresh your compute and your storage is fine? Well you have to buy new storage anyway. Plus, it's a lot easier budget-wise to buy compute one year, then refresh the SAN a couple years later.
     
    • Like
    Reactions: sryan2k1 and SandyTech
    S

    sryan2k1

    Ars Legatus Legionis
    44,493
    Subscriptor++
  • #38
    HCI has never made sense to me unless you had a very uniform workload and/or idiots running it.


    ESXi, a iSCSI SAN, and some Ethernet switches isn't rocket science.

    While Nutanix seems to be the best, there are so many others that are a dumpster fire. VxRail is usually referred to as VxFail for a reason. "Single throat to choke"? Nope! Everyone is going to blame everyone.


    I don't remember if it was Nutanix or another but someone I know was burned because their cluster was EOL and to add another node the entire cluster had to be forklifted.
     
    Paladin

    Paladin

    Ars Legatus Legionis
    32,552
    Subscriptor
  • #40
    Yeah I think the concept of HCI is nice in a lot of cases but the reality is that it almost always works out more expensive, less flexible, and not even that much easier to use, support, or upgrade/build. It doesn't really leave behind a lot of the problems of more traditional deployments, mostly because a lot of them come from business decisions and needs rather than technical ones and those business issues come from people, not hardware or software.

    That said, there is a market for an affordable HCI type deployment that has a good design to provide some redundancy and path for growth. The problem is you have a very hard time finding that. It always feels like there is a big cliff to climb for the first deployment compared to a single server setup and then an even bigger cliff to climb if you outgrow or outlive the normal 'small' deployment of an HCI system. Once you need to upgrade or update it, you're back to a big migration job or looking at a new platform because the one you had is end of life or never had a good upgrade path or something.
     
    You must log in or register to reply here.
    Jump to latest Follow Reply
    Jump to latest Follow Reply