I am trying to work out whether it is practical to propose Dom0 xen support as a feature for Fedora 15. The kernel situation is that Domain 0 has been accepted upstream for 2.6.37. Assuming a 3 month kernel release cycle, F15 will most likely ship with a 2.6.37.x kernel, with 2.6.38 coming out either after the F15 release or just before but too late to be included. If the plan to get key xen drivers into 2.6.38 succeeds, then F15 may be become usable as a Domain 0 system at some point during its lifetime as the kernel package in a Fedora version typically has one major update. If the kernel team accept backported patches then it might just be possible to ship F15 with usable Domain 0 support but the timescale for that would be very tight. The other thing we would need to consider is what needs to be done to make xen friendly enough to be usable by an ordinary user. The page https://fedoraproject.org/wiki/Features/XenPvopsDom0 contains plans from when dom0 xen support was expected to make a quick return to Fedora, but they are a couple of years old now so probably need updating. I think as a minimum we would need a way to add a dom0 enabled grub entry for a kernel, rather than requiring the user to hand edit the grub file. We should also make sure that xen works with the other Fedora virtualisation tools. What do others think about this? For example is it achievable as a feature, is it too early and better to wait for F16, and what else should we aim to do to make xen usable in Fedora? Michael Young

If the kernel team accept backported patches then it might just be possible to ship F15 with usable Domain 0 support but the timescale for that would be very tight.

I am trying to work out whether it is practical to propose Dom0 xen support as a feature for Fedora 15. The kernel situation is that Domain 0 has been accepted upstream for 2.6.37. Assuming a 3 month kernel release cycle, F15 will most likely ship with a 2.6.37.x kernel, with 2.6.38 coming out either after the F15 release or just before but too late to be included. If the plan to get key xen drivers into 2.6.38 succeeds, then F15 may be become usable as a Domain 0 system at some point during its lifetime as the kernel package in a Fedora version typically has one major update. If the kernel team accept backported patches then it might just be possible to ship F15 with usable Domain 0 support but the timescale for that would be very tight. The other thing we would need to consider is what needs to be done to make xen friendly enough to be usable by an ordinary user. The page https://fedoraproject.org/wiki/Features/XenPvopsDom0 contains plans from when dom0 xen support was expected to make a quick return to Fedora, but they are a couple of years old now so probably need updating. I think as a minimum we would need a way to add a dom0 enabled grub entry for a kernel, rather than requiring the user to hand edit the grub file. We should also make sure that xen works with the other Fedora virtualisation tools.

I am trying to work out whether it is practical to propose Dom0 xen support as a feature for Fedora 15. The kernel situation is that Domain 0 has been accepted upstream for 2.6.37. Assuming a 3 month kernel release cycle, F15 will most likely ship with a 2.6.37.x kernel, with 2.6.38 coming out either after the F15 release or just before but too late to be included. If the plan to get key xen drivers into 2.6.38 succeeds, then F15 may be become usable as a Domain 0 system at some point during its lifetime as the kernel package in a Fedora version typically has one major update. If the kernel team accept backported patches then it might just be possible to ship F15 with usable Domain 0 support but the timescale for that would be very tight. The other thing we would need to consider is what needs to be done to make xen friendly enough to be usable by an ordinary user. The page https://fedoraproject.org/wiki/Features/XenPvopsDom0 contains plans from when dom0 xen support was expected to make a quick return to Fedora, but they are a couple of years old now so probably need updating. I think as a minimum we would need a way to add a dom0 enabled grub entry for a kernel, rather than requiring the user to hand edit the grub file. We should also make sure that xen works with the other Fedora virtualisation tools. What do others think about this? For example is it achievable as a feature, is it too early and better to wait for F16, and what else should we aim to do to make xen usable in Fedora? Michael Young _______________________________________________ virt mailing list virt(a)lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/virt

On Mon, 8 Nov 2010, Don Dutile wrote: > IMO, the pieces that the Xen folks have to whip into shape are the tools; > the missing kernel pieces are fairly well known (backend drivers), although Xen folks > are considering blkback in userspace instead of kernel space, so that may be > a new twist. I did see a post which mentioned that qemu had blkback and netback and wondering if they would work with xen, but I don't know any more than that.

On 11/08/2010 04:55 AM, M A Young wrote: > I am trying to work out whether it is practical to propose Dom0 xen > support as a feature for Fedora 15. > > The kernel situation is that Domain 0 has been accepted upstream for > 2.6.37. Assuming a 3 month kernel release cycle, F15 will most likely ship > with a 2.6.37.x kernel, with 2.6.38 coming out either after the F15 > release or just before but too late to be included. If the plan to get key > xen drivers into 2.6.38 succeeds, then F15 may be become usable as a > Domain 0 system at some point during its lifetime as the kernel package in > a Fedora version typically has one major update. > > If the kernel team accept backported patches then it might just be > possible to ship F15 with usable Domain 0 support but the timescale for > that would be very tight. > > The other thing we would need to consider is what needs to be done to make > xen friendly enough to be usable by an ordinary user. The page > https://fedoraproject.org/wiki/Features/XenPvopsDom0 contains plans from > when dom0 xen support was expected to make a quick return to Fedora, but > they are a couple of years old now so probably need updating. > > I think as a minimum we would need a way to add a dom0 enabled grub entry > for a kernel, rather than requiring the user to hand edit the grub file. > We should also make sure that xen works with the other Fedora > virtualisation tools. > > What do others think about this? For example is it achievable as a > feature, is it too early and better to wait for F16, and what else should > we aim to do to make xen usable in Fedora? > Have you consider kvm? it's upstream since 2.6.20 and now its more ready than ever.

----- Original Message ----- > From: "Bill Davidsen"<davidsen(a)tmr.com> > To: dlaor(a)redhat.com > Cc: xen(a)lists.fedoraproject.org, virt(a)lists.fedoraproject.org, "M A Young"<m.a.young(a)durham.ac.uk> > Sent: Monday, November 8, 2010 11:52:08 AM > Subject: Re: [fedora-virt] [Fedora-xen] Dom0 xen support in Fedora 15? > Dor Laor wrote: >> On 11/08/2010 04:55 AM, M A Young wrote: >> >>> I am trying to work out whether it is practical to propose Dom0 xen >>> support as a feature for Fedora 15. >>> >>> The kernel situation is that Domain 0 has been accepted upstream >>> for >>> 2.6.37. Assuming a 3 month kernel release cycle, F15 will most >>> likely ship >>> with a 2.6.37.x kernel, with 2.6.38 coming out either after the F15 >>> release or just before but too late to be included. If the plan to >>> get key >>> xen drivers into 2.6.38 succeeds, then F15 may be become usable as >>> a >>> Domain 0 system at some point during its lifetime as the kernel >>> package in >>> a Fedora version typically has one major update. >>> >>> If the kernel team accept backported patches then it might just be >>> possible to ship F15 with usable Domain 0 support but the timescale >>> for >>> that would be very tight. >>> >>> The other thing we would need to consider is what needs to be done >>> to make >>> xen friendly enough to be usable by an ordinary user. The page >>> https://fedoraproject.org/wiki/Features/XenPvopsDom0 contains plans >>> from >>> when dom0 xen support was expected to make a quick return to >>> Fedora, but >>> they are a couple of years old now so probably need updating. >>> >>> I think as a minimum we would need a way to add a dom0 enabled grub >>> entry >>> for a kernel, rather than requiring the user to hand edit the grub >>> file. >>> We should also make sure that xen works with the other Fedora >>> virtualisation tools. >>> >>> What do others think about this? For example is it achievable as a >>> feature, is it too early and better to wait for F16, and what else >>> should >>> we aim to do to make xen usable in Fedora? >>> >> Have you consider kvm? it's upstream since 2.6.20 and now its more >> ready >> than ever. >> > > There are some good tutorials which should explain the difference > between xen and kvm, particularly the performance and hardware > requirements of each. re: hardware requirements, KVM's requirement for VT-X/AMD-V extensions certainly used to be a concern 2-3 years ago but today even laptops come with this support. And regarding performance they days of Xen outperforming KVM have long-since passed.

----- Original Message ----- > From: "Dennis Jacobfeuerborn" <dennisml(a)conversis.de&gt; > To: "Andrew Cathrow" <acathrow(a)redhat.com&gt; > Cc: "Bill Davidsen" <davidsen(a)tmr.com&gt;, xen(a)lists.fedoraproject.org, virt(a)lists.fedoraproject.org, "M A Young" > <m.a.young(a)durham.ac.uk&gt; > Sent: Monday, November 8, 2010 1:59:54 PM > Subject: Re: [fedora-virt] [Fedora-xen] Dom0 xen support in Fedora 15? > On 11/08/2010 06:02 PM, Andrew Cathrow wrote: > > > > > > ----- Original Message ----- > >> From: "Bill Davidsen&quot;&lt;davidsen(a)tmr.com&gt; > >> To: dlaor(a)redhat.com > >> Cc: xen(a)lists.fedoraproject.org, virt(a)lists.fedoraproject.org, "M A > >> Young&quot;&lt;m.a.young(a)durham.ac.uk&gt; > >> Sent: Monday, November 8, 2010 11:52:08 AM > >> Subject: Re: [fedora-virt] [Fedora-xen] Dom0 xen support in Fedora > >> 15? > >> Dor Laor wrote: > >>> On 11/08/2010 04:55 AM, M A Young wrote: > >>> > >>>> I am trying to work out whether it is practical to propose Dom0 > >>>> xen > >>>> support as a feature for Fedora 15. > >>>> > >>>> The kernel situation is that Domain 0 has been accepted upstream > >>>> for > >>>> 2.6.37. Assuming a 3 month kernel release cycle, F15 will most > >>>> likely ship > >>>> with a 2.6.37.x kernel, with 2.6.38 coming out either after the > >>>> F15 > >>>> release or just before but too late to be included. If the plan > >>>> to > >>>> get key > >>>> xen drivers into 2.6.38 succeeds, then F15 may be become usable > >>>> as > >>>> a > >>>> Domain 0 system at some point during its lifetime as the kernel > >>>> package in > >>>> a Fedora version typically has one major update. > >>>> > >>>> If the kernel team accept backported patches then it might just > >>>> be > >>>> possible to ship F15 with usable Domain 0 support but the > >>>> timescale > >>>> for > >>>> that would be very tight. > >>>> > >>>> The other thing we would need to consider is what needs to be > >>>> done > >>>> to make > >>>> xen friendly enough to be usable by an ordinary user. The page > >>>> https://fedoraproject.org/wiki/Features/XenPvopsDom0 contains > >>>> plans > >>>> from > >>>> when dom0 xen support was expected to make a quick return to > >>>> Fedora, but > >>>> they are a couple of years old now so probably need updating. > >>>> > >>>> I think as a minimum we would need a way to add a dom0 enabled > >>>> grub > >>>> entry > >>>> for a kernel, rather than requiring the user to hand edit the > >>>> grub > >>>> file. > >>>> We should also make sure that xen works with the other Fedora > >>>> virtualisation tools. > >>>> > >>>> What do others think about this? For example is it achievable as > >>>> a > >>>> feature, is it too early and better to wait for F16, and what > >>>> else > >>>> should > >>>> we aim to do to make xen usable in Fedora? > >>>> > >>> Have you consider kvm? it's upstream since 2.6.20 and now its more > >>> ready > >>> than ever. > >>> > >> > >> There are some good tutorials which should explain the difference > >> between xen and kvm, particularly the performance and hardware > >> requirements of each. > > > > re: hardware requirements, KVM's requirement for VT-X/AMD-V > > extensions certainly used to be a concern 2-3 years ago but today > > even laptops come with this support. > > And regarding performance they days of Xen outperforming KVM have > > long-since passed. > > Citations needed. I'm not saying what you claim isn't true but without > data > this opinion doesn't carry much weight. > Citations are really needed on both sides of the debate, 2 or 3 year old metrics no longer apply. Vendors published benchmarks are typically questionable, they focus on their products strengths and their competitors weakness. The only hope for a fair comparison is a vendor neutral set of benchmarks such as SPECvirt http://www.spec.org/virt_sc2010/ But obviously this isn't a simple test to run. > Regards, > Dennis -- xen mailing list xen(a)lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/xen

AFAIK there are some issues with that benchmark not being free. What I'm basically looking for is a simple comparison of a few key metrics between common setups. For example a simple bonnie++ run on XenPV and KVM+virtio_blk could already give people at least some baseline numbers. Regards, Dennis

Digimer, I would think the latest xen vs the latest kvm would be best. Fully virtualized vs Para-virtualized, is that possible with kvm? I don't believe your specs are too low end, they actually match pretty close to what I have and I don't think the OS matters that much as long as they are identical setups. Thanks for offering to do the comparison, I am very interested in your results. Jon

On 10-11-09 12:54 PM, jonr(a)destar.net wrote: > Digimer, > > I would think the latest xen vs the latest kvm would be best. Fully > virtualized vs Para-virtualized, is that possible with kvm? I don't > believe your specs are too low end, they actually match pretty close > to what I have and I don't think the OS matters that much as long as > they are identical setups. > > Thanks for offering to do the comparison, I am very interested in your > results. > > Jon I will look into whether fully virtualized is an option for KVM. If not, I probably won't worry as all "modern" operating systems support paravirtualized setups.

I've got no way of knowing how to run tests on non-linux VMs. If anyone can give advice on benchmarking non-Linux OS', I'd be grateful and will give it a go. Can you (or anyone) suggests tests to run beyond bonnie++ and a kernel compile that would be "real-world" and useful benchmarks? I think one test that will help would be to run tests concurrently on two VMs to see if there is a difference in how Xen or KVM handle very random disk I/O. PS - I will likely need a week or so to get these tests done. -- Digimer E-Mail: digimer(a)alteeve.com AN!Whitepapers: http://alteeve.com Node Assassin:  http://nodeassassin.org -- xen mailing list xen(a)lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/xen

Digimer wrote: > On 10-11-09 11:49 AM, Dennis Jacobfeuerborn wrote: >> AFAIK there are some issues with that benchmark not being free. >> >> What I'm basically looking for is a simple comparison of a few key metrics >> between common setups. For example a simple bonnie++ run on XenPV and >> KVM+virtio_blk could already give people at least some baseline numbers. >> >> Regards, >> Dennis > I'd be happy to run some benchmarks. I've got some identical (if humble) > machines... Would you want just a Xen vs. KVM? If so, what parameters or > setup options would provide the best apples to apples comparison? > > The machines are; > - quad core athlon II x4 > - 4gb ddr3 > - 1x 7200rom 500GB drive > - Fedora 14, stripped (but identical) installs > > If these are too low end for useful results, let me know. Otherwise, > I'll fire up Xen under Michael's 2.6.32-25-172 kernel with the stock > 4.0.1 hypervisor on one machine, and can use the stock KVM/QEMU setup > for another. > > Assuming all is fine, would it matter whether the VMs were F14 vs CentOS > 5.5? > > I was thinking 3x runs each of; > > - bonnie++ > - Recompile Michael's kernel > - ?? Please make suggestions > Thanks for the offer. The services which have in the past shown significant differences (my experience) are small, with lots of connects and io. Examples are NNTP (ak usenet news) and POP mail servers. I suspect that DHCP and DNS have too little io to show the effect, but I never had enough load to know. Unless web service has a lot of dynamic content, I doubt it would have enough load to show the difference unless someone has server image and load generator to use. I could probably dig up a simulated load for NNTP, but I don't have an easy way to preload a server. My days of generating test loads were when I ran ISP NNTP servers and only had to provide test loading against the servers.

Performance aside, There is still a lot of old or small hardware that just can't do KVM at all, and which are easier to administer with a VM which can be moved.

> I'd be happy to run some benchmarks. I've got some identical (if humble) > machines... Would you want just a Xen vs. KVM? If so, what parameters or > setup options would provide the best apples to apples comparison? > > The machines are; > - quad core athlon II x4 > - 4gb ddr3 > - 1x 7200rom 500GB drive > - Fedora 14, stripped (but identical) installs > > If these are too low end for useful results, let me know. Otherwise, > I'll fire up Xen under Michael's 2.6.32-25-172 kernel with the stock > 4.0.1 hypervisor on one machine, and can use the stock KVM/QEMU setup > for another. > > Assuming all is fine, would it matter whether the VMs were F14 vs CentOS > 5.5? > > I was thinking 3x runs each of; > > - bonnie++ > - Recompile Michael's kernel > - ?? Please make suggestions There is also an ongoing discussion about Xen vs. KVM performance on the CentOS virt list http://lists.centos.org/pipermail/centos-virt/2010-October/002074.html

On 10-11-12 09:57 AM, Todd Deshane wrote: > There is also an ongoing discussion about Xen vs. KVM performance on > the CentOS virt list > > http://lists.centos.org/pipermail/centos-virt/2010-October/002074.html There are some interesting ideas and insightful comments there. I am tempted to abort my own benchmarking, as it seems others are doing the same. I think I will proceed though, until and unless someone there releases numbers first. As I mentioned before, I've hit a snag with my in-progress benchmarks. I've filed a bug but I doubt I will wait for resolution. I'll need a bit more time now to setup a RHEL 6 based platform (host and guests). I'm still hoping to get most done this weekend, but I make no promises. If not, I'll aim for the weekend after to finish. Cheers

On Mon, Nov 08, 2010 at 11:52:08AM -0500, Bill Davidsen wrote: > Dor Laor wrote: > >> On 11/08/2010 04:55 AM, M A Young wrote: >> >>> What do others think about this? For example is it achievable as a >>> feature, is it too early and better to wait for F16, and what else should >>> we aim to do to make xen usable in Fedora? >>> >>> >> Have you consider kvm? it's upstream since 2.6.20 and now its more ready >> than ever. >> >> > There are some good tutorials which should explain the difference > between xen and kvm, particularly the performance and hardware > requirements of each. > In any case, the question of whether KVM or Xen is best, is not really relevant to whether Xen Dom0 has a place in F15. Fedora will welcome any software that meets the packaging& licensing guidelines, and has someone who is willing to maintain it. So if people want to maintain Xen as an alternative virtualization option in Fedora, they're welcome todo so. KVM will of course remain the default virt host setup offered in the installer

> And I would hope that the installer would be clever enough to see if KVM > is supported and offer xen if not. Doesn't need to be default, available > is fine. What if i don't want KVM install no matter of hardware ? I will want Xen on icore7 + ASUS P6TDT+ 16 GB RAM , Dell PowerEdge and so on ...

On Tue, Nov 09, 2010 at 10:40:10AM -0800, Boris Derzhavets wrote: > > And I would hope that the installer would be clever enough to see if KVM > > is supported and offer xen if not. Doesn't need to be default, available > > is fine. > > What if i don't want KVM install no matter of hardware ? > I will want Xen on icore7 + ASUS P6TDT+ 16 GB RAM , Dell PowerEdge and so on ... I guess you will have to manually yum install kernel-xen.  We can't magically know your preference, but it makes sense to default to KVM (by which I am really saying "default to a plain Linux kernel") because that is far less intrusive for users.

> And I would hope that the installer would be clever enough to see if KVM > is supported and offer xen if not. Doesn't need to be default, available > is fine. What if i don't want KVM install no matter of hardware ? I will want Xen on icore7 + ASUS P6TDT+ 16 GB RAM , Dell PowerEdge and so on ...

I am trying to work out whether it is practical to propose Dom0 xen support as a feature for Fedora 15. The kernel situation is that Domain 0 has been accepted upstream for 2.6.37. Assuming a 3 month kernel release cycle, F15 will most likely ship with a 2.6.37.x kernel, with 2.6.38 coming out either after the F15 release or just before but too late to be included. If the plan to get key xen drivers into 2.6.38 succeeds, then F15 may be become usable as a Domain 0 system at some point during its lifetime as the kernel package in a Fedora version typically has one major update. If the kernel team accept backported patches then it might just be possible to ship F15 with usable Domain 0 support but the timescale for that would be very tight. The other thing we would need to consider is what needs to be done to make xen friendly enough to be usable by an ordinary user. The page https://fedoraproject.org/wiki/Features/XenPvopsDom0 contains plans from when dom0 xen support was expected to make a quick return to Fedora, but they are a couple of years old now so probably need updating. I think as a minimum we would need a way to add a dom0 enabled grub entry for a kernel, rather than requiring the user to hand edit the grub file. We should also make sure that xen works with the other Fedora virtualisation tools. What do others think about this? For example is it achievable as a feature, is it too early and better to wait for F16, and what else should we aim to do to make xen usable in Fedora? Michael Young

On 11/07/2010 03:55 PM, M A Young wrote: > If the plan to get key > xen drivers into 2.6.38 succeeds With the caveat that I don't really know what I'm looking for, I didn't see these make 2.6.38-rc1 (plenty of other Xen work, though). I saw a proposed patch for netback hit xen-devel yesterday. So, did we miss Fedora 15?

On 11/07/2010 03:55 PM, M A Young wrote: > If the plan to get key > xen drivers into 2.6.38 succeeds With the caveat that I don't really know what I'm looking for, I didn't see these make 2.6.38-rc1 (plenty of other Xen work, though). I saw a proposed patch for netback hit xen-devel yesterday.

On Fri, Jan 21, 2011 at 08:39:46AM +0000, M A Young wrote: > On Thu, 20 Jan 2011, Bill McGonigle wrote: > > > On 11/07/2010 03:55 PM, M A Young wrote: > >> If the plan to get key > >> xen drivers into 2.6.38 succeeds > > > > With the caveat that I don't really know what I'm looking for, I didn't see > > these make 2.6.38-rc1 (plenty of other Xen work, though). I saw a proposed > > patch for netback hit xen-devel yesterday. > > There was talk of userland netback and blkback in xen 4.1 which use the > gntdev interface which did make it into 2.6.38 (userland blkback > apparently gives the similar performance as the kernel one, netback is > not as good). I am not sure of the details though, nor how you might use > them if they are available. Xen folks worked to give QEMU a userland netback/blkback implementation, but I don't believe this functionality is integrated into the mgmt stack at this time, so likely not much use yet.