Comments on: A comparison of important features between Citrix XenServer 5, Microsoft Hyper-V 1.0 and ESXi 3.5u3 http://blog.core-it.com.au/?p=311 Mon, 30 Aug 2010 04:38:29 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: Ron kuper http://blog.core-it.com.au/?p=311&cpage=1#comment-973 Ron kuper Tue, 10 Mar 2009 14:06:37 +0000 http://blog.core-it.com.au/?p=311#comment-973 Flexibility is not a bad thing. That something is technically possible (e.g. using non WHQL Drivers) is not to say that it is good or recommended. I'm not ruling out VMWare's drivers from being optimized (are they really making their own drivers for all of the qualified hardware?), but if a vendor can and allow you to use uncertified/3rd party drivers does not necessarily mean that the certified drivers are not doing a great job. I don't think that we can currently say which approach is better (Control domain vs. drivers in the hypervisor). Both have pros and cons. For instance with the driver domain you can have several drivers domain and have each guest connected through a different domain so that a driver fault will not bring down the host. Just don't degrade the discussion to "Optimized" vs. "Generic". Doing so means ignoring important points. Flexibility is not a bad thing.

That something is technically possible (e.g. using non WHQL Drivers) is not to say that it is good or recommended.

I’m not ruling out VMWare’s drivers from being optimized (are they really making their own drivers for all of the qualified hardware?), but if a vendor can and allow you to use uncertified/3rd party drivers does not necessarily mean that the certified drivers are not doing a great job.

I don’t think that we can currently say which approach is better (Control domain vs. drivers in the hypervisor). Both have pros and cons. For instance with the driver domain you can have several drivers domain and have each guest connected through a different domain so that a driver fault will not bring down the host.

Just don’t degrade the discussion to “Optimized” vs. “Generic”. Doing so means ignoring important points.

]]> By: Leo http://blog.core-it.com.au/?p=311&cpage=1#comment-971 Leo Tue, 10 Mar 2009 08:58:26 +0000 http://blog.core-it.com.au/?p=311#comment-971 Depends. See, with Hyper-V, the nature of the OS means you can use any NIC and any HBA, whether or not it has been tested in WHQL labs. This means that drivers are naturally less hardened and less controlled than the VMware ones where VMware works with the IHVs to develop drivers specifically for ESX/ESXi. Depends. See, with Hyper-V, the nature of the OS means you can use any NIC and any HBA, whether or not it has been tested in WHQL labs.

This means that drivers are naturally less hardened and less controlled than the VMware ones where VMware works with the IHVs to develop drivers specifically for ESX/ESXi.

]]> By: Ron kuper http://blog.core-it.com.au/?p=311&cpage=1#comment-967 Ron kuper Mon, 09 Mar 2009 11:00:44 +0000 http://blog.core-it.com.au/?p=311#comment-967 Control Domain is not necessarily a disadvantage as you imply in the table ("Optimized Drivers vs. Generic Drivers"). "Optimized" is a bit misleading. Can you say linux/windows drivers as less "Optimized" than VMware's? Drivers in the hypervisior/Drivers in a "Control Domain" are different hypervisor architectures (such as full emulation vs. paravirtualization), each with its own pros and cons. Benchmarks are available and favor one platform to another usually depending on the type of load tested. (And, unfortunately, the affiliation of the person who tested it). Every organization which considers virtualization and cares about performance should perform its own scalability and performance testing. Anyway "Optimized" have no real value as useful information. (And much less so "Generic"…) Control Domain is not necessarily a disadvantage as you imply in the table (“Optimized Drivers vs. Generic Drivers”).

“Optimized” is a bit misleading. Can you say linux/windows drivers as less “Optimized” than VMware’s?

Drivers in the hypervisior/Drivers in a “Control Domain” are different hypervisor architectures (such as full emulation vs. paravirtualization), each with its own pros and cons.
Benchmarks are available and favor one platform to another usually depending on the type of load tested. (And, unfortunately, the affiliation of the person who tested it).
Every organization which considers virtualization and cares
about performance should perform its own scalability and performance testing.

Anyway “Optimized” have no real value as useful information.
(And much less so “Generic”…)

]]> By: Leo http://blog.core-it.com.au/?p=311&cpage=1#comment-945 Leo Tue, 27 Jan 2009 01:31:53 +0000 http://blog.core-it.com.au/?p=311#comment-945 Hi Andrew, You're right on the size - ESXi installs in 560MB - my mistake. I've also dropped VMsafe - when I made the table originally I had it on good authority that the release was not far away. There is a host OS for XenServer and Hyper-V - neither can run without a control domain to buffer I/O and network info. While the same is true with ESXi, it's not a control domain per se. This is reasonably easy to see - Xen utilises Linux drivers for all devices, as does Server Core. ESX/ESXi have Linux-derived drivers but they are not loaded in the Linux instance. Rather they are loaded into the vmkernel. Cheers, Leo Hi Andrew,

You’re right on the size – ESXi installs in 560MB – my mistake. I’ve also dropped VMsafe – when I made the table originally I had it on good authority that the release was not far away.

There is a host OS for XenServer and Hyper-V – neither can run without a control domain to buffer I/O and network info.

While the same is true with ESXi, it’s not a control domain per se. This is reasonably easy to see – Xen utilises Linux drivers for all devices, as does Server Core. ESX/ESXi have Linux-derived drivers but they are not loaded in the Linux instance. Rather they are loaded into the vmkernel.

Cheers,
Leo

]]> By: Andrew http://blog.core-it.com.au/?p=311&cpage=1#comment-943 Andrew Mon, 26 Jan 2009 07:48:17 +0000 http://blog.core-it.com.au/?p=311#comment-943 I think the size on disk is a bit misleading - While it's true that XenServer requires 16GB to install, that's not the size of the hypervisor - which is what you refer to with ESXi. If you want to compare the hypervisor size, then 32 MB for ESXi, 500kb XenServer (I think Hyper-v is 1.5 mb but not sure). Please amend your table with the required disk space to install each product or the hypervisor size, but at least make it consistent! I'd drop the security row as well - VMSafe is currently vapourware and has been for about 4 years. maybe it might get released with VI4, but not yet... I could go on and on (there is no host OS for XenServer or Hyper-v - both are bare metal hypervisors which utilise a control domain - even ESXi has a control domain which is Busybox Linux). I think the size on disk is a bit misleading – While it’s true that XenServer requires 16GB to install, that’s not the size of the hypervisor – which is what you refer to with ESXi.
If you want to compare the hypervisor size, then 32 MB for ESXi, 500kb XenServer (I think Hyper-v is 1.5 mb but not sure).
Please amend your table with the required disk space to install each product or the hypervisor size, but at least make it consistent!
I’d drop the security row as well – VMSafe is currently vapourware and has been for about 4 years. maybe it might get released with VI4, but not yet…
I could go on and on (there is no host OS for XenServer or Hyper-v – both are bare metal hypervisors which utilise a control domain – even ESXi has a control domain which is Busybox Linux).

]]> By: Comparison: Citrix XenServer 5, Microsoft Hyper-V 1.0 and ESXi 3.5 « Hotware: Dirk’s Software Blog http://blog.core-it.com.au/?p=311&cpage=1#comment-897 Comparison: Citrix XenServer 5, Microsoft Hyper-V 1.0 and ESXi 3.5 « Hotware: Dirk’s Software Blog Mon, 08 Dec 2008 18:08:57 +0000 http://blog.core-it.com.au/?p=311#comment-897 [...] A comparison of important features between Citrix XenServer 5, Microsoft Hyper-V 1.0 and ESXi 3.5 [...] [...] A comparison of important features between Citrix XenServer 5, Microsoft Hyper-V 1.0 and ESXi 3.5 [...]

]]> By: A comparison of important features between Citrix XenServer 5, Microsoft Hyper-V 1.0 and ESXi 3.5u3 | Richard Parmiter http://blog.core-it.com.au/?p=311&cpage=1#comment-857 A comparison of important features between Citrix XenServer 5, Microsoft Hyper-V 1.0 and ESXi 3.5u3 | Richard Parmiter Tue, 25 Nov 2008 09:31:37 +0000 http://blog.core-it.com.au/?p=311#comment-857 [...] Originally posted here [...] [...] Originally posted here [...]

]]> By: Leo http://blog.core-it.com.au/?p=311&cpage=1#comment-839 Leo Mon, 17 Nov 2008 04:41:40 +0000 http://blog.core-it.com.au/?p=311#comment-839 Hi Gabrie, Yes, it's true that you can find a system but the fact of the matter is that I'm running Xen and Hyper-V on custom-built white boxes on the cheapest components I could find. You can't do that with VMware. You can only match what VMware ahs drivers for. Cheers, Leo Hi Gabrie,

Yes, it’s true that you can find a system but the fact of the matter is that I’m running Xen and Hyper-V on custom-built white boxes on the cheapest components I could find.

You can’t do that with VMware. You can only match what VMware ahs drivers for.

Cheers,
Leo

]]> By: Gabrie http://blog.core-it.com.au/?p=311&cpage=1#comment-838 Gabrie Sun, 16 Nov 2008 16:28:55 +0000 http://blog.core-it.com.au/?p=311#comment-838 Hi Regarding the last point (Hardware comptibility), it is true that ESX has a strict HCL, but when you look at the large number of supported systems, you will surely find a system that is supported and suited for your needs. Hi
Regarding the last point (Hardware comptibility), it is true that ESX has a strict HCL, but when you look at the large number of supported systems, you will surely find a system that is supported and suited for your needs.

]]> By: Leo http://blog.core-it.com.au/?p=311&cpage=1#comment-837 Leo Fri, 14 Nov 2008 23:04:05 +0000 http://blog.core-it.com.au/?p=311#comment-837 Hi Tamas, I've updated - thank you. Hi Tamas,

I’ve updated – thank you.

]]>