All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Yes this topic again: Xen, KVM or OpenVZ
AnthonySmith
Member, Patron Provider
Hi Folks,
I know I know this has been done to death but I wanted a more recent take on things from end users while I appreciate the LET/B user base is not the whole world it is a good representation and if nothing else fairly honest about things 
The reason I am asking this is over the last few months I have been having a bit of a mental struggle with continuing with Xen, it is and will always be my own preference however with the obvious lack of support in terms of automation for the hosting industry I am starting to feel a little cornered in terms of future proofing, the Xen development in solusvm has been stale at best for at least 3 years and progress in the hosting industry seems to be centered around OpenVZ and KVM.
So I would appreciate opinions on the following:
1) Based on nothing else but the virt type would you select Xen or KVM or OpenVZ?
2) Based on nothing else but the virt type would you select Xen or KVM
3) Have you ever avoided a service just because it was Xen and if so why?
4) Are you aware Xen HVM supports installation from ISO just like KVM?
5) If you are an Inception Hosting customer already and the option to convert your current Xen VPS (HVM or PV) to KVM would you want to? (would include data)
If anyone just has any general thoughts and does not want to answer the questions that is fine, it is not a loaded topic
Anthony.
Comments
It really depends, you can't pick one really. Some unuses might warrant KVM over Xen, but with me it's always Xen-PV or KVM.
If I had to choose, KVM, but they're pretty much equal in my mind and KVM only swings it down to it's rep. Xen is perfectly acceptable to me too.
^^
No, my production box atm is Xen
Yes, but I've never done it. Just seen it on google when I wanted to play with BSD.
I'm not other than LES but in general I'd probably stick with what I have if it's the same price, location etc. Don't fix what isn't broken and that. Unless it was being phased out.
Hope it helps!
In sequence OpenVZ -> Xen -> KVM you have more control and less speed.
State the task, then choose proper virtualization.
I will never buy OpenVZ anymore, no matter what plan/location/price. It just sucks way too much.
KVM, absolutely.
Not really, if other aspects are attractive for some reason, I could try living with Xen.
A tiny minority of hosts use Xen HVM, it's practically safe to assume when you see "Xen", it's Xen PV, and HVM is either not an option, or costs extra.
I do not have active Inception Hosting services at the moment, but if a Xen host that I use asked me if I want to swich over to KVM with everything else being the same, then I'd be more than happy to.
Benefits of KVM over Xen-PV are mostly in the boot-up and crash recovery aspect:
1) a KVM VPS uses its own bootloader, which can be just standard normal GRUB
2) KVM provides a proper VNC console, so you can connect early in the boot-up cycle and see why your VPS fails to boot, or select a different boot option within said bootloader. Not sure if it's possible to provide a VNC console on Xen-PV, but even if it is, the majority of Xen-PV hosts don't provide it.
So I had numerous issues on Xen, where after a kernel update the VPS would not boot anymore. And since I had no VNC access, the only option was to ask the support to mount my VPS image and edit a file in it, so that the default boot option becomes the previous kernel. In one case the problem was that the kludge called "py-grub" that goes for a boot loader on Xen PV failed to recognize an XZ-compressed kernel. In another, I don't even know what, and have no way of knowing, just rolled back to an old kernel and stayed on that one until I cancelled the VPS altogether.
Most importantly, Xen-PV does not provide any benefit to me as the end user to justify tolerating all that bootloader b/s. KVM with all virtio drivers is not even any measurably slower. The only case I'll buy Xen is if the offer is cheap and specs or location are good. Even then, each kernel update is basically a game of chance to hose up your VPS and have to wait possibly multiple hours for a support reply, so would be wary of hosting anything critical on it.
I do not consider a comparison between XEN PV and KVM exists. More along the lines of XEN PV and OpenVZ. In my eyes the comparison rests between XEN HVM and KVM.
I have servers with Inception Hosting because of the XEN HVM support. I choose XEN for stability.
Although a bit dated: (http://www-archive.xenproject.org/files/xensummitboston08/Deshane-XenSummit08-Slides.pdf) shows some intertesting stats. I find the most interesting is the scalability (XEN scales with N complexity, KVM not so much). Unfortunately there are more studies for XEN than KVM - due to enterprise support for XEN. Other studies support XEN for stability, while I see none for KVM.
The performance of the two virtualizations is similar as both support hardware passthrough. I saw higher stats for KVM in disk IO, but higher stats for XEN in network IO. Both were similar enough. Therefore I went with XEN for the stability alone.
I will only switch to KVM if there was a motivation to offset my perception of XEN stability.
Thanks @rm_ your thoughts essentially mirror my own.
@Silvenga back end comparison, yes indeed I agree and my personal preference will always be Xen, the reason for this question is more around requirements for providing a scale-able product to end users under webhosting conditions with current automation availability.
Just for clarity under no circumstances am I even considering a forced migration from Xen to KVM, that would never end well
Xen PV or Xen HVM? I'd choose Xen PV, but KVM if HVM is the alternate.
Ditto
I've avoided an Xen HVM vps before.
Are you aware for linux distributions that support xen that xen pv can install from ISO almost just like kvm?
If I was, no.
Xen HVM then KVM and OpenVZ last
Thanks @Rallias regarding point 4, I am aware however to my knowledge no commercially available panels that are fit for the hosting industry support it.
1) Xen
2) Xen
3) n/a
4) Yes
5) no
Yeah, true, but that's because it's a total utter ballache to pull off. I use it personally to boot systemrescuecd on my lab machine.
I can write a book about it. All abusive cores and sugar vps clients are on xen hvm. You can create ISO repo, then do net boot for that specific ISO. That is how we do it. So basically you are using PXE boot with custom ISO. But in that case, good bye solusVM, virtualizor and all of those nice panels... You will have to go with something custom or buy provisioning system for Dedicated servers. We have solved problem thanks to @rds100 as he pointed us with NOC-PS. We made some custom modules for it but we are now able to provision whatever you wish that is bootable.
I understand I perhaps do not have a realistic view on this in terms of understanding the scope of effort/knowledge required as it is not my area of experience but I often (very often) wonder why nothing exists yet that actually genuinely works and covers everything.
We have panels for: OpenVZ, Xen, KVM, VMware/Esx, LXC, hell even virtualbox but none of them do everything and to my knowledge none of them do a particularly great job of anything.
Let me put this to you as I know you do have some experience in this area:
You are given 2 years of full time paid work and a development infrastructure and a budget for third party service buy in such as GUI design/ graphics/ pen testing to develop the following:
1) A web based panel for administration and end user management (with the hosting industry in mind) for Xen/KVM/OpenVZ
2) A built in billing module and also modules for whmcs/hostbill/(the other one)
3) A feature set then genuinely reflected the capabilities of each technology not just the basics
4) A very well written migration module that covers those 3 virt types and also the billing (whmcs/hostbill/etc) elements.
Is that even remotely achievable for 1 person and if not what is with 2 years of full time paid work? or how many years or people would it actually take?
I know this is finger to the wind stuff just looking for an opinion, it just genuinely seems like madness to me that no one else can even compete (not intended as a solusvm put down) with solusvm, and not just that but nothing even competes with the version of solusvm that was available 4 years ago.
But Xen HVM supports install direct from ISO with solusvm out of the box so either I am missing your point (quite possible) or you missed mine?
Oh sorry, just looked up NOC-PS I understand what you are saying now
good prices, shame they did a hostbill and even blesta module but not WHMCS, again seems like commercial madness.
Well... everything except billing module... I've just built the billing management into the panel itself.
Honestly, it would take a good developer. You can't just go to a development house in india that just shovels out code like there's no tomorrow and expect a quality, you have to get an experienced developer, but they should be able to shovel out a working prototype in 3 months for all of it, have it nice, flush, secure, and beautiful within about 6. But it is reasonably possible.
I mean... just look at Feathur. One person (and for part of the time a technical consultant) was able to put out a product that far exceeds SolusVM's capability set securely, even exceeding it's capability in the realms of KVM and OpenVZ in that it has code to do things SolusVM DREAMS of. Granted, if I had my way, a few things would be done differently about it, but I don't have my way.
All the commercial panels support ISO for HVM.
I missed yours, sorry for this. It is quite normaly and possible to instal custom iso from solus. We are using part of hostbill part of our custom panel. Integration lever into billing panel (hostibill) and xenserver was superbad. So i am completely in my story... i didn't consider that you are using whmcs and solus.
Sorry for this
Yes. Assuming full time work is 30-35+ hours a week this is certainly achievable. Maybe with one person you'd have a bit of a tough time having all three (four) big players supported thoroughly and well tested but it is defiantly possible.
As I'm sure you know as a web host, the panels don't necessarily do a lot of independent work if you know what I mean, they just tie together other services, edit config files, etc.
It is one of those how long is a piece of string kinds of things, but on the surface it's possible, definitely seeing as you'd be willing to outsource some parts.
I guess finding someone to work solo on a first party product for two years is a different ballgame though. It's a lot of work for one person if you can't resell your work again.
there is whmcs module ready for production. Old but working...
https://www.noc-ps.com/downloads_download_free_trial/
Interesting, I would never use a software house for anything like this, the reason I asked is that I am actually in the second phase/ stage of securing grant funding available to business in my area for creating jobs through innovative software/ technology and approach and this is essentially the tiny synopses of what I am applying for funding for.
I fully expect I will be flying someone in/ relocating someone though but long term I can justify how it will create jobs.
Sadly in the mean time I am stuck with sucking it up and living with what is available.
I just literally CANNOT believe there is no viable alternative.
I bet an Open Source development team could get this done in a year. There just lacks the motivation to do so.
Well, I'm probably not the dev you'd need, but if whoever you do choose needs technical consultancy, I am extremely well versed in the intrinsics of all the technologies involved.
I was thinking 18 months to get something released and then 6 months for improvements and marketing.
My feeling is a lot of people end up at a stage of to much effort/ potential issues with migrating from one product to another and to much risk to outlay for bespoke products with no commercial return.
I will use featur is a microcosmic example I hope Justin does not mind
, seems like a great product it does what it says in the tin so to speak but immediately prevents anyone with Xen services from migrating, the migration afaik has no automation and the billing migration has to be done manually and lets face it no one wants to sit and update 6000+ client records manually and have potential issues due to possible problems with manual migrations with 1000's of users/ services.
If an alternative existed that had already considered these things in the core development I would genuinely bet money that if is is an improvement at no extra cost that brings more features to end users and has a seamless migration process a lot of the bigger hosts would jump at the chance.
Yep a few per hour consultants would be key.
i didn't try Xen but would choose KVM over openVZ.
@drserver reply from NOC-PS when asking about WHMCS module future development: "We do not see any future in WHMCS."
Again another short sighted view IMO given its popularity and significantly higher user base over the 2 they do support.
KVM and OpenVZ are kind of on a par for me, OpenVZ because of the ease of upgrading and changing stuff such as disk and ram, and KVM because of the isolation from the host, ability to use custom stuff and the ability to boot into a live rescue CD over VNC and backup the whole VPS without needing the hosts help.
Q: Based on nothing else but the virt type would you select Xen or KVM or OpenVZ?
A: Xen
Q: Based on nothing else but the virt type would you select Xen or KVM
A: Xen
Q: Have you ever avoided a service just because it was Xen and if so why?
A: No
Q: Are you aware Xen HVM supports installation from ISO just like KVM?
A: Yes.
Q: If you are an Inception Hosting customer already and the option to convert your current Xen VPS (HVM or PV) to KVM would you want to? (would include data)
A: Yes I am, and I would, if I had no other option.
Personally, I'm a KVM guy. I'd choose Xen HVM and KVM before OpenVZ though.
1) Based on nothing else but the virt type would you select Xen or KVM or OpenVZ?
Most projects fit nicely with OpenVZ as long as I get a VPS from a good provider (BuyVM, Hostigation, Iniz, LES, Ramnode, by alphabet order) it depends on how well they monitor the node.
If the budget allows it, I get a KVM better resource managment and I can tweak a lot more things.
If Xen has the same price as OpenVZ, Xen is my choice.
2) Based on nothing else but the virt type would you select Xen or KVM?
KVM
3) Have you ever avoided a service just because it was Xen and if so why?
No, it just was a bit more difficult to load a custom kernel (prgmr.com) it was my first experience setting up a custom kernel on Xen
4) Are you aware Xen HVM supports installation from ISO just like KVM?
Yes, but some Windows apps on Xen HVM dont work and on KVM they work well.
5) If you are an Inception Hosting customer already and the option to convert your current Xen VPS (HVM or PV) to KVM would you want to? (would include data)
For the same price, I would be getting an account (dont have a Xen box with you at the moment).