Broadcom's VMware buy got you worried? Give these 5 FOSS hypervisors a spin Last month we shared four VMware ESXi alternatives for enterprises hedging their bets over Broadcom's impending takeover of the virtualization giant. You can find those suggestions and all your comments right here. But for small to midsized businesses looking for an escape from VMware's stranglehold on virtualization, you may …
"Another option, which is neither free nor open source, but then neither is VM-Ware, is Microsoft's Hyper-V."
Hyper-V Server *is* a free baremetal hypervisor, though not open source. It is a cut down version of Windows Server with only the virtualization components - no GUI or even Explorer, just Powershell.
To use GUI, you need to manage it with any Windows Workstation / Server computer, or for heavy usage you probably would need to invest into SCVMM.
Hyper-V seems to be going away at the moment, with Micros~1 peddling the baremetal Azure Stack HCI successor which requires licensing.
The last version of the 'free' Hyper-V server was 2019, so it would not be advisable to start moving to it now as there won't be any more versions.
As you say, they are moving to Azure Stack HCI - which is far more tightly integrated with Microsoft's cloudy services, requires at least two hosts, and you have to pay monthly licensing fees to run your VMs in addition to any licensing fees for the OSs running on the VMs.
I've seen companies issue urgent "remove all VirtualBox instances now" warnings, time and time again over the past several years. Turns out, Oracle loves to find companies using such, try and force an audit on them that will cost the company a lot of money even if they are in compliance, and if they find even one user who without thinking is using it commercially in a manner only licensed for personal use, hit the company with huge penalties. It was literally cheaper to buy alternatives for users than to go through such audits for a "free" software product.
I've seen companies issue urgent "remove all VirtualBox instances now" warnings, time and time again over the past several years.
It's the Extension Pack that is not FOSS. Oracle logs IP addresses when you download something then, when they've logged enough downloads of the Extension Pack or Java 8u202 or above or whatever, start making enquiries to your company about licensing.
Innotek VirtualBox: Awesome. Oracle VirtualBox: Avoid!
Oracle took a great product and turned it into shite. The VBox programmers maniacally prioritize "cool" new features at the expense of code quality. I had looked at their buglist and found regression bugs which deleted users' data! If programmers are doing things "properly", each error fixed will have a corresponding automated test to ensure future fixes and feature additions don't re-introduce that bug. Oracle's VBox programmers are clearly not doing that, and those sorts of (to me) mal-priorities are a consequence of the corporate culture there. (Those mal-priorities / bad cultures can and do exist in non-corporate FOSS projects as well.)
VirtualBox depends on a ton of closed-source plugins and does not deserve a mention in an article about FOSS virtualization. QEMU and KVM are better in their own right and I wish I would have dropped VirtualBox at least 5 years earlier than I did.
VirtualBox has one closed source plug-in that you can ignore if your VM doesn't need USB 2/3 support. The basic product is GPL2 and perfectly usable as-is.
Oracle's licence for the plug-in is quite restrictive and their pricing (minimum $5000) explains why businesses are keen not to get caught with a bootleg copy on someone's machine, but it doesn't make the rest of the product worthless.
At the end of the day it's the classic, what are the SLAs on your production workloads?
If you're need 5-nines min, then you got few choices but to go with VMWare or cloud compute resources. However if your needs are far simpler and you might only need to keep minor prod loads running during business day, an hour of batches at closing time and all done until 9am next morning a lot of these, even Hyper-Vm VirtualBox and Q-EMU could well suffice if you have people on hand to look after it and a little down time every so often is no biggie.
I have Veeam backups nicely integrated with vSphere. Jenkins is nicely integrated with vSphere as well. Replacing that will take time and effort. And that's our development lab setup.
Moreover we sell a product that works on vSphere as well - so we have need dogfooding unless we also replace VMWare there - which would be not little effort as well, and customers already running on VMWare would have their issues too.
I really do hope Broadcom doesn't screw up everything, because replacing VMWare will be a bloodbath.
Make hay while the sun shines.
But particularly that customer point - I'd thoroughly consider looking at migration paths off for your customers as an option. Even if vmware remains decent there might be a few forced off and it might suck to lose them!
Your own struggles I'm sure you could overcome if and when you need to quickly enough.
The problem is the product is tightly integrate with VMWare (including NSX and other features). First we would have to rewrite the application to work on another virtualization system keeping the actual features, than we would have to migrate customers deployment. And customers have to retrain their VMWare administrators, and also modify their backup/monitorig/etc. systems as well.
Easier to day than doing it - especially now we have a tight budget because of the actual situation.
Yeah. I can imagine not knowing the product it could absolutely be a nightmare. And I'm not suggesting forcing customers off against their will or anything!
Just highlighting that if I was procuring a product, right now with the uncertainty, tied to vmware I'd seriously be considering alternative options.
Similarly if I was a current customer and licensing/support from the vmware side was my responsibility and not a part of the product/service. I'd be asking what alternative upgrade paths there were for future!
You have hit the nail on the head here.
If you have VMware or Hyper-v then pretty much every mainstream backup product will integrate seamlessly.
Vendors are not going to offer support for a myriad of different hypervisors that are all edge cases. Switching back to in-os backup agents is not the answer either.
I'm sorry but this is a pretty poor article. While Proxmox, Harvester and XCP-NG are indeed valid alternatives to ESXi, listing a management plane (OpenNebula) and desktop hypervisor (Virtualbox? seriously?) in the same breath is, frankly, silly.
What a shame as there there are other alternatives that could have made the list, such as oVirt Node, FreeBSD bhyve or simply KVM+cockpit.
Ovirt was a pig to install for me, I had to end up hacking some of the ansible install scripts to get it to actually install OK. The documentation is also very sparse (and close to useless). But once up and running it's OK.
Red Hat Enterprise Virtualisation (effectively the same thing with a RH badge of for those that don't know) installed fine. Again seems to be OK when running. But I hear RH are changing their offering around virtualisation to be more with OpenShift.
I think the article was good but ‘replace VMWare’ is quite an ambiguous statement!
The hypervisor layer, for a Windows shop the choice is easy enough - Hyper-V, until the estate is more than a few boxes and then the ‘grown up’ VM manager (often called ‘the control plane software’) is pricey. It’s all very Windows-y with the need to use powershell on the cli being almost unavoidable.
Linux - KVM is a good choice - it supports running several different guest operating systems (such as Windows). Handy if you want a Linux VM on an existing Linux host. KVM is a Linux kernel module so it’s very convenient for a developer with a few development hosts, but it’s not (on its own) ‘just’ a hypervisor. However, there a a ton of companies selling what is essentially a distro which includes the components to run KVM and not much else, which is branded the ‘hypervisor product’. Generally, you need a vendo’s hyper visor product in order to integrate easily into the control plane s/w which is the part where you’ll need to open your wallet. There are lots of good control plane options.
Xen-ng - is great hypervisor but Xen-orchestra ’free’ doesn’t include the ability to schedule full VM backups, so the right license will be required for anyone managing other folk’s VMs and that’ll cost you (a lot less than vCenter!!).
QEMU is an emulator, not a hypervisor - and last time I used it, a pain to get working well. In practice - I can’t think of a good reason I’d consider QEMU if I was starting a new project. I certainly wouldn’t consider it as VMWare replacement as it isn’t in the same functional realm as ESXi - because it’s a different thing.
Hope that helps answer your questions.
> QEMU is an emulator
QEMU is used with KVM it's how all the IO is done.
It's been the best part of decade but I seem to remember that is also fully virtuallized XEN worked.
QEMU can be used to emulate different models of CPU.
Before you could rely on the processors have the virtualization extensions QEMU could run where KVM couldn't.
But most QEMU does the bits that KVM sees as an SEP.
Hello James,
1. It's XCP-ng, not Xen-ng ;)
2. Xen Orchestra is entirely free/Open Source. You can have all features by just install it from the sources on Github. This is in the official doc! https://xen-orchestra.com/docs/installation.html#from-the-sources What you pay is the turnkey virtual appliance with pro support :)
My answer: there are lots of tools, if you go for a paid for hypervisor then the tooling will be supplied by the vendor. If not, you’ll need to use one of the several available FOSS utilities, but you first need to export your ESXi VM in EXACTLY the right format for the tool to be able to do it’s thing. Things like guest HDD file format, secure boot, and guest BIOS/UFEI seem to cause headaches for many of the conversation tools. Even when the tool outputs an image you’ll find things like the network interfaces may get messed up. These are a problems that OVF is intended to improve, but we’re not there yet! Once you’ve got a good set of steps, it’s easy. Finding the right set of steps is a little time consuming, and changes over time, and careful testing of the end VM will be required.
For most moving off of VMware to anything other than cloud based hosting would be very difficult due to the number of systems that integrate in such as Backups, which those of us that actually do manage this stuff have to maintain. Bear in mind that if you keep your backups for any length of time, such years, you will need to maintain a recovery environment, personally we only keep 3 months so not the end of the world but would need to find a product comparable to Veeam for any of these solutions.
Also we are using HCI with VSAN so no separate storage, we would therefore need to set up all that stuff as well, would be expensive, as you don't mention anything that is comparable in this gloss over. We also use NSX and virtual networking is also not mentioned so may be addition cost add on and complexity to move.
Hyper-v is free for most people as they probably already have per core Datacenter licences on their hosts for Windows, unless you are heavy in to Linux guests then this is probably the way most would go.
Having spent almost a year migrating 500 servers from an NSX-v platform to an NSX-t one and juggling the business requirements, other projects' needs and down time etc on what was relatively simple process, moving to another completely different platform would not be joyful or quick experience.
To me this is a case of Broadcom buying something at the peak of its value, most people in medium to large sized businesses are moving to hybrid hosting ahead of a move to cloud, solutions that offer this as a reasonable simple migration path are going to be the winners especially as hardware starts to get replaced and if Broadcom do a Broadcom and stuff it up, Microsofts Azure Stack HCI seems to offer this kind of path so would be a likely first choice for many even if like most Register readers you hate Microsoft they may just have the smoothest path when you have to deal with all those technology people that are not technology people in the business.
Microsofts Azure Stack HCI has many drawbacks though - it's far from a drop-in VMware replacement. The hardware must be maintained absolutely identical between nodes and maxes out at 16 nodes. It has a single egress IP for instances and the alternative VPN option is Windows-specific and very slow. You need to shut the entire platform down just to add RAM or a GPU. You can't manage multiple stacks via a single control plane, so managing multi-site is a nightmare (although I believe they are working on that).
The equivalent of VSAN for Proxmox would be Ceph, which Proxmox can deploy and manage for you. Ceph can be a bit complex to learn (Proxmox helps here!) but actually very reliable once it's running.
I'd also echo the above drawbacks regarding about Azure Stack. It's got a familiar UI and API to those already using Azure but that's really the only strong card it has in its hand.
How is VirtualBox licensed?
The VirtualBox base package contains the full VirtualBox source code and platform binaries and is licensed under the GNU General Public License, version 2. You can distribute and modify the base package, provided that you distribute all modifications under the GPLv2 as well.
The VirtualBox Extension Pack is available under the VirtualBox Extension Pack Personal Use and Evaluation License, which is a free license for personal, educational or evaluation use, or an Enterprise License, which is a for-fee license that allows most commercial, non-distribution uses restricted by the PUEL.
Background: I am a Solutions Architect for Red Hat. I work with many customers from small to large enterprises. I have worked with many organizations that have expressed a desire to move away from VMWare.
First, a point I would like to highlight. There is a difference between Hypervisors (eg: ESXi, KVM, Hyper-V, Xen) and "Virtualization Platforms" (eg: RHV/oVirt, Proxmox, vSphere, OpenStack). Hypervisor is an underlying technology that implements virtualization, whereas the platform manages and integrates that hypervisor with automation, monitoring, scaling, networking, resource management, etc.) Almost all businesses need a Virtualization Platform, not just a Hypervisor.
Of the organizations I have worked with, most use VMWare, some use Red Hat Virtualization (the Product that for which oVirt is the upstream development project, and which uses QEMU/KVM under the hood), and others use OpenStack, HyperV, Proxmox, Nutanix and more. Some organizations use multiple solutions.
Having worked with all of these organizations for many years, here is my observation: With the exception of very small deployments, it is exceptionally difficult to create a compelling business case to change hypervisor platforms. It is like changing the engines of an airplane in-flight. The longer an organization uses a technology, the more its operational process becomes integrated and dependent on that technology. Many here have already pointed out integration with backups. But there is also integration with networking, ITSM systems, automation, and applications. There is training, operational experience, and knowledge in how those systems work, and their ties into other parts of the business.
The magic isn't in the Hypervisor (though, frankly, the VMWare Hypervisor is very good), it's in the systems that manage and integrate that hypervisor, such as vSphere, NSX, vRA/vRO, etc.
So, alas, most of these 'alternatives' aren't really alternatives for anyone other than a hobbyist or small business that doesn't have to manage VMs at scale. I would say Proxmox is the most likely option for anyone small. There are some large deployments of Red Hat Virtualization (oVirt), and it's a pretty solid enterprise virtualization platform. As others are noted, Red Hat's focus is shifting to OpenShift Virtualization (based on the upstream KubeVirt project), so the oVirt community would need an infusion of non-Red Hat contributors to get wind back in its sails. KubeVirt-based solutions aren't yet ready for many of the use-cases that are currently filled by VMWare. In fact, some of those use-cases may never be a good fit for container-based virtualization, as the approaches are not the same, even though the underlying hypervisor is.
With all that said, what do I recommend to organizations that want my opinion? Don't think about your virtualization platform in isolation. Think about your business, your apps, and what value you get for what you pay. If your virtualization platform costs a lot of money, but also delivers a lot of benefit, and the cost of switching it out doesn't give you a business advantage, why do it? On the other hand, if your business is being hampered by mountains of technological debt, you have poor reliability, or other challenges, maybe it makes sense to plan a future that takes advantage of newer technology, uses more products that use an open-source development model, etc. Don't run away from something you think is bad, plot a course to something better.
But whatever you do, don't just 'lift and shift' your VM's to the cloud because of a belief that it will 'save you money'. That is almost never true in the long run.
If users are using VMware vDirector with vShpere, a much better solution than those listed that involve exporting VMs is CloudStack. CloudStack provides an approach of cataloging existing VMware vShpere virtual assets without having to convert the instances to another hypervisor. So, without much fuss, after the resource re-catalog process, users can add other hypervisors to the same infrastructure and remove resources in an organized and smooth way.
I have kicked the tires on this entire stack, and the hands down winner is ProxMox PVE 8.1.3. It shames all others on a performance basis, will host VM's for OS's from every flavor of Linux, Windows, and some Mac flavours, and is scalable on a cluster level that meets most needs without licensing of any sort.
Pull down the ISO here and give yourself a break from the licensing disease.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
