Just reporting some iperf results. In each case, Dom0 is iperf server, DomU is iperf client: (1) Dom0: Intel Core2 3.16 GHz, CentOS 5.2, xen 3.0.3. DomU: Windows XP SP3, GPLPV 0.9.12-pre13, file based. Iperf: 1.17 Gbits/sec (2) Dom0: Intel Core2 2.33 GHz, CentOS 5.2, xen 3.0.3. DomU: Windows XP SP3, GPLPV 0.9.12-pre13, file based. Iperf: 725 Mbits/sec (3) Dom0: Intel Core2 2.33 GHz, CentOS 5.2, xen 3.0.3. DomU: Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. Iperf: 1.03 Gbits/sec Outstanding. Steve _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Anno domini 2009 Steve Thompson scripsit: Hi!> Just reporting some iperf results. In each case, Dom0 is iperf server, > DomU is iperf client:> (1) Dom0: Intel Core2 3.16 GHz, CentOS 5.2, xen 3.0.3. > DomU: Windows XP SP3, GPLPV 0.9.12-pre13, file based. > Iperf: 1.17 Gbits/sec> (2) Dom0: Intel Core2 2.33 GHz, CentOS 5.2, xen 3.0.3. > DomU: Windows XP SP3, GPLPV 0.9.12-pre13, file based. > Iperf: 725 Mbits/sec> (3) Dom0: Intel Core2 2.33 GHz, CentOS 5.2, xen 3.0.3. > DomU: Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. > Iperf: 1.03 Gbits/sec> Outstanding.:) Is there any chance, you could try this again with a more recent Xen version? I would be really interested if there is any difference. Thanks for providing the values. Ciao Max -- Follow the white penguin. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sat, 17 Jan 2009, Maximilian Wilhelm wrote:> Is there any chance, you could try this again with a more recent Xen > version? I would be really interested if there is any difference.Yes, I can do that, but it will take some time. There are 33 DomU''s in production on this host; I''ll see about a test rig. Steve _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
jeroen groenewegen van der weyden
2009-Jan-18 09:15 UTC
Re: [Xen-users] GPLPV network performance
gentlemen can I join this game, Ik made these tests also, but I don''t see by for the performance you mentioned below? is there something I forgot to tweak? Dom0: Intel Q9450 2.66 GHz, OpenSuse11.0 64bit, xen 3.2.1_16881 DomU: Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. Iperf: 547 Mbits/sec (domu -> dom0) Best regards Jeroen. Steve Thompson wrote:> On Sat, 17 Jan 2009, Maximilian Wilhelm wrote: > >> Is there any chance, you could try this again with a more recent Xen >> version? I would be really interested if there is any difference. > > Yes, I can do that, but it will take some time. There are 33 DomU''s in > production on this host; I''ll see about a test rig. > > Steve > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users > ------------------------------------------------------------------------ > > > No virus found in this incoming message. > Checked by AVG - http://www.avg.com > Version: 8.0.176 / Virus Database: 270.10.8/1899 - Release Date: 17-1-2009 17:50 > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> > gentlemen can I join this game, Ik made these tests also, but I don''t > see by for the performance you mentioned below? is there something I > forgot to tweak? > > Dom0: Intel Q9450 2.66 GHz, OpenSuse11.0 64bit, xen 3.2.1_16881 > DomU: Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. > Iperf: 547 Mbits/sec (domu -> dom0) >Also need to know: . have you disabled any of the offload features in the windows network driver? (if you don''t know how to do that then your answer is no :) . how many vcpu''s does Windows have? . Is your windows 2003 32 or 64 bit? James _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sun, Jan 18, 2009 at 6:57 PM, James Harper <james.harper@bendigoit.com.au> wrote:>> is there something I >> forgot to tweak? >> >> Dom0: Intel Q9450 2.66 GHz, OpenSuse11.0 64bit, xen 3.2.1_16881 >> DomU: Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. >> Iperf: 547 Mbits/sec (domu -> dom0) >> > > Also need to know: > . have you disabled any of the offload features in the windows network > driver? (if you don''t know how to do that then your answer is no :) > . how many vcpu''s does Windows have? > . Is your windows 2003 32 or 64 bit? >Last week I setup two Windows 2003 32bit (one of them is cloned from the other), GPLPV, 2 CPUs each, with ACPI=1 on domU config. iperf got around 500 Mbps. Then out of curiosity, I changed one of them to have one CPU, ACPI=0, and changed Windows HAL to Standard PC. iperf now gives me a whopping 1.4 Gbps. Both tests where done without messing with any network offload settings. Is it safe to assume that, for now, Standard PC HAL will give better network performance compared to ACPI HAL when used with GPLPV? Is it also the same case for the disk driver? Regards, Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
jeroen groenewegen van der weyden
2009-Jan-18 13:38 UTC
Re: [Xen-users] GPLPV network performance
. have you disabled any of the offload features in the windows network driver? (if you don''t know how to do that then your answer is no :) NO . how many vcpu''s does Windows have? 1 vcpu . Is your windows 2003 32 or 64 bit? 32 bit best regards, jeroen James Harper wrote:>> gentlemen can I join this game, Ik made these tests also, but I don''t >> see by for the performance you mentioned below? is there something I >> forgot to tweak? >> >> Dom0: Intel Q9450 2.66 GHz, OpenSuse11.0 64bit, xen 3.2.1_16881 >> DomU: Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. >> Iperf: 547 Mbits/sec (domu -> dom0) >> >> > > Also need to know: > . have you disabled any of the offload features in the windows network > driver? (if you don''t know how to do that then your answer is no :) > . how many vcpu''s does Windows have? > . Is your windows 2003 32 or 64 bit? > > James > ------------------------------------------------------------------------ > > > No virus found in this incoming message. > Checked by AVG - http://www.avg.com > Version: 8.0.176 / Virus Database: 270.10.8/1899 - Release Date: 17-1-2009 17:50 > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> On Sun, Jan 18, 2009 at 6:57 PM, James Harper > <james.harper@bendigoit.com.au> wrote: > >> is there something I > >> forgot to tweak? > >> > >> Dom0: Intel Q9450 2.66 GHz, OpenSuse11.0 64bit, xen 3.2.1_16881 > >> DomU: Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. > >> Iperf: 547 Mbits/sec (domu -> dom0) > >> > > > > Also need to know: > > . have you disabled any of the offload features in the windowsnetwork> > driver? (if you don''t know how to do that then your answer is no :) > > . how many vcpu''s does Windows have? > > . Is your windows 2003 32 or 64 bit? > > > > > Last week I setup two Windows 2003 32bit (one of them is cloned from > the other), GPLPV, 2 CPUs each, with ACPI=1 on domU config. iperf got > around 500 Mbps. > > Then out of curiosity, I changed one of them to have one CPU, ACPI=0, > and changed Windows HAL to Standard PC. iperf now gives me a whopping > 1.4 Gbps. > > Both tests where done without messing with any network offload > settings. Is it safe to assume that, for now, Standard PC HAL will > give better network performance compared to ACPI HAL when used with > GPLPV? >That is a good question, but I don''t know the answer. I get great performance (>2Gbits/second) on my system without much effort at all, even with SMP. My system is AMD though... maybe there is some difference there? With 1 cpu and ACPI, what was your performance like? (eg was it SMP slowing you down or ACPI?) I would be grateful if you could do a bit more testing here as this is a recurring problem...> Is it also the same case for the disk driver?Probably. I think the problem is system based rather than simply network based. James _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
James Harper wrote:>> Is it safe to assume that, for now, Standard PC HAL will >> give better network performance compared to ACPI HAL when used with >> GPLPV? >> >> > > That is a good question, but I don''t know the answer. I get great > performance (>2Gbits/second) on my system without much effort at all, > even with SMP. My system is AMD though... maybe there is some difference > there? > >I''m using Intel on RHEL 5.2, bundled Xen. Maybe that''s what causing it.> With 1 cpu and ACPI, what was your performance like? (eg was it SMP > slowing you down or ACPI?) > >The thing is I can''t. Note easily, anyway. I wrote in another thread (GPLPV drivers problem), changing "vcpus=1" on the SMP domU crashed it. Changing to HAL=halaacpi doesn''t help either.> I would be grateful if you could do a bit more testing here as this is a > recurring problem... > >I guess I could try installing a fresh 1-CPU system. This might take several hours though. Thanks for your help. Regards, Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hi James, Fajar A. Nugraha wrote:> James Harper wrote: > >>> Is it safe to assume that, for now, Standard PC HAL will >>> give better network performance compared to ACPI HAL when used with >>> GPLPV? >>> >>> >>> >> That is a good question, but I don''t know the answer. I get great >> performance (>2Gbits/second) on my system without much effort at all, >>By "without much effort" you mean not having to change any network offload settings, right?>> even with SMP. My system is AMD though... maybe there is some difference >> there? >> >> >> > > I''m using Intel on RHEL 5.2, bundled Xen. Maybe that''s what causing it. > >>> With 1 cpu and ACPI, what was your performance like? (eg was it SMP >> slowing you down or ACPI?) >> >>I tested again using: - Xeon X5450 @3GHz (same machine as earlier test), W2k3, single CPU, ACPI -> iperf = 1.4 Gbps - Opteron 2220 SE @2.8 GHz, W2k3, single CPU, ACPI -> iperf = 1 Gbps So it seems that AMD or Intel doesn''t matter much, and neither does ACPI vs standard PC. The only thing that makes HUGE difference is SMP. Is this a known issue? Regards, Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sun, Jan 18, 2009 at 9:26 PM, Fajar A. Nugraha <fajar@fajar.net> wrote:> Hi James, > > F > > > I tested again using: > - Xeon X5450 @3GHz (same machine as earlier test), W2k3, single CPU, > ACPI -> iperf = 1.4 Gbps > - Opteron 2220 SE @2.8 GHz, W2k3, single CPU, ACPI -> iperf = 1 Gbps > > So it seems that AMD or Intel doesn''t matter much, and neither does ACPI > vs standard PC. The only thing that makes HUGE difference is SMP. Is > this a known issue? > > Regards, > > Fajar >I''d be curious about DomU to DomU performance. Nobody seems to test this very often but if you have a database server or file server in a DomU that''s serving out to another then DomU to DomU performance would be critical. Grant McWilliams Some people, when confronted with a problem, think "I know, I''ll use Windows." Now they have two problems. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Grant McWilliams wrote:> On Sun, Jan 18, 2009 at 9:26 PM, Fajar A. Nugraha <fajar@fajar.net > <mailto:fajar@fajar.net>> wrote: > > I tested again using: > - Xeon X5450 @3GHz (same machine as earlier test), W2k3, single CPU, > ACPI -> iperf = 1.4 Gbps > - Opteron 2220 SE @2.8 GHz, W2k3, single CPU, ACPI -> iperf = 1 Gbps > > > I''d be curious about DomU to DomU performance. Nobody seems to test > this very often but if you have a database server or file server in a > DomU that''s serving out to another then DomU to DomU performance would > be critical. > >Hmm. I just tested domU <-> domU on the same Xeon dom0 (same config as above), and got 530 Mbps. The domU that functions as iperf server uses 100% CPU from xm top (both domUs are using different physical CPU). Oh well, at least it''s well above 100 Mbps. Regards, Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
I only get 155Mbit/sec DomU to DomU regardless of driver settings. CPU on Server hits 100%, client runs at 33%. System is Opteron 8350 based, one fixed CPU core per VM on Xen 3.3.0/Centos 5.2, Pre13 GPLPV, CPUID fix applied. Turning on Receive moderation make the score worse. Do the GPLPV drivers still have general performance issues with Xen 3.3.0 or is this just my setup? Thanks Rob -----Original Message----- From: xen-users-bounces@lists.xensource.com [mailto:xen-users-bounces@lists.xensource.com] On Behalf Of Fajar A. Nugraha Sent: 19 January 2009 09:27 To: xen-users Subject: Re: [Xen-users] GPLPV network performance Grant McWilliams wrote:> On Sun, Jan 18, 2009 at 9:26 PM, Fajar A. Nugraha <fajar@fajar.net > <mailto:fajar@fajar.net>> wrote: > > I tested again using: > - Xeon X5450 @3GHz (same machine as earlier test), W2k3, single CPU, > ACPI -> iperf = 1.4 Gbps > - Opteron 2220 SE @2.8 GHz, W2k3, single CPU, ACPI -> iperf = 1 > Gbps > > > I''d be curious about DomU to DomU performance. Nobody seems to test > this very often but if you have a database server or file server in a > DomU that''s serving out to another then DomU to DomU performance would > be critical. > >Hmm. I just tested domU <-> domU on the same Xeon dom0 (same config as above), and got 530 Mbps. The domU that functions as iperf server uses 100% CPU from xm top (both domUs are using different physical CPU). Oh well, at least it''s well above 100 Mbps. Regards, Fajar The SAQ Group Registered Office: 18 Chapel Street, Petersfield, Hampshire GU32 3DZ SAQ is the trading name of SEMTEC Limited. Registered in England & Wales Company Number: 06481952 http://www.saqnet.co.uk AS29219 SAQ Group Delivers high quality, honestly priced communication and I.T. services to UK Business. Broadband : Domains : Email : Hosting : CoLo : Servers : Racks : Transit : Backups : Managed Networks : Remote Support. ISPA Member Find us in http://www.thebestof.co.uk/petersfield _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Sun, 18 Jan 2009, Grant McWilliams wrote:> I''d be curious about DomU to DomU performance. Nobody seems to test this > very often but if you have a database server or file server in a DomU that''s > serving out to another then DomU to DomU performance would be critical.Previously I had obtained these numbers for a Dom0 of Intel Core2 2.33 GHz, CentOS 5.2, xen 3.0.3: Windows XP SP3, GPLPV 0.9.12-pre13, file based. 725 Mbits/sec Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. 1.03 Gbits/sec For DomU(XP) to Domu(2003), I get 462 Mbits/sec, about in line with what I would have expected. The Dom0 was the same machine for both DomU''s; I assume that the throughput would be higher if they were different. I didn''t note at the time, but the xen version of 3.0.3 applies only to the xen utilities package itself; from "xm dmesg", it shows a version of 3.1.2-92.1.22.el5. Steve _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon, Jan 19, 2009 at 6:57 AM, Steve Thompson <smt@vgersoft.com> wrote:> On Sun, 18 Jan 2009, Grant McWilliams wrote: > > I''d be curious about DomU to DomU performance. Nobody seems to test this >> very often but if you have a database server or file server in a DomU >> that''s >> serving out to another then DomU to DomU performance would be critical. >> > > Previously I had obtained these numbers for a Dom0 of Intel Core2 2.33 GHz, > CentOS 5.2, xen 3.0.3: > > Windows XP SP3, GPLPV 0.9.12-pre13, file based. 725 Mbits/sec > Windows 2003 SP2, GPLPV 0.9.12-pre13, file based. 1.03 Gbits/sec > > For DomU(XP) to Domu(2003), I get 462 Mbits/sec, about in line with what I > would have expected. The Dom0 was the same machine for both DomU''s; I assume > that the throughput would be higher if they were different. > > I didn''t note at the time, but the xen version of 3.0.3 applies only to > the xen utilities package itself; from "xm dmesg", it shows a version > of 3.1.2-92.1.22.el5. > > Steve >These numbers are what I expect but irritate me to no end. I don''t just want to have Virtual Servers that provide services to the outside world (which is usually a connection a great deal slower than 1 Gbit) but rather provide services to other DomUs and in this case where it would be beneficial to have a super fast network it isn''t. It''s actually slower than if you were to provide services to the outside world. It seems the speed issues are with data entering the DomU and not actually DomU to DomU performance. Can you guys test Dom0 to DomU speed instead of the other way around. I suspect you will get 1Gbit from DomU to Dom0 but about half that going the other way from Dom0 to DomU. Grant McWilliams Some people, when confronted with a problem, think "I know, I''ll use Windows." Now they have two problems. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon, 19 Jan 2009, Grant McWilliams wrote:> It seems the speed issues are with data entering the DomU and not > actually DomU to DomU performance. Can you guys test Dom0 to DomU speed > instead of the other way around.Dom0 2.33 GHz Intel Core2, CentOS 5.2, xen 3.0.3, GPLPV 0.9.12-pre13: Windows XP SP3: server=Dom0 client=DomU 702 Mbits/sec server=DomU client=Dom0 667 Mbits/sec (5% slower) Windows 2003 SP2: server=Dom0 client=DomU 1.02 Gbits/sec server=DomU client=Dom0 719 Mbits/sec (29% slower) Steve _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> On Mon, 19 Jan 2009, Grant McWilliams wrote: > > > It seems the speed issues are with data entering the DomU and not > > actually DomU to DomU performance. Can you guys test Dom0 to DomUspeed> > instead of the other way around. > > Dom0 2.33 GHz Intel Core2, CentOS 5.2, xen 3.0.3, GPLPV 0.9.12-pre13: > > Windows XP SP3: > server=Dom0 client=DomU 702 Mbits/sec > server=DomU client=Dom0 667 Mbits/sec (5% slower) > > Windows 2003 SP2: > server=Dom0 client=DomU 1.02 Gbits/sec > server=DomU client=Dom0 719 Mbits/sec (29% slower)Just FYI, the reason for the performance drop in the domU RX path is that Windows doesn''t support some of the offload functions that Linux does, so I have to emulate them which has some pretty high overhead, mostly related to packet copying. In fact in the RX path, there is quite a lot of copying going on and there is definitely room for improvement, but it will never be as good as TX. James _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> > I only get 155Mbit/sec DomU to DomU regardless of driver settings. CPUon> Server hits 100%, client runs at 33%. System is Opteron 8350 based,one> fixed CPU core per VM on Xen 3.3.0/Centos 5.2, Pre13 GPLPV, CPUID fix > applied. Turning on Receive moderation make the score worse. Do theGPLPV> drivers still have general performance issues with Xen 3.3.0 or isthis> just my setup? >DomU->DomU isn''t a scenario I have tested recently... I just tested it then and am getting around 350MBit/second from XP->2K3 DomU''s, vs 2GBit/second XP->Dom0. The latest upload (0.9.12-pre15-dont-use) includes a windows patch for XP which you can enable with /PATCHTPR in your boot.ini, which will give a substantial boost to XP which may explain the difference in performance between my testing (with /PATCHTPR) and yours. If you do want to test 0.9.12-pre15-dont-use with the /PATCHTPR boot.ini option, please do so on a pure test box. I haven''t had any problems but I have only tested with XPsp2 and XPsp3. When I first tested 3.3.0 I definitely noticed a large performance drop from 3.2.x, but I never did enough testing to blame 3.3.0 with any certainty. I use 3.3.1 now and don''t have any performance issues in DomU<->Dom0 network throughput. Those DomU->DomU figures (mine and yours) are obviously less than desirable though, given the DomU->Dom0 performance... I''ll see what I can do. James _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tue, 20 Jan 2009, James Harper wrote:> Those DomU->DomU figures (mine and yours) are obviously less than > desirable though, given the DomU->Dom0 performance... I''ll see what I > can do.James, I would just like to raise my hand and join the sea of hands; you have done a kick-ass job with this. Thank you! Steve _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon, Jan 19, 2009 at 3:50 PM, James Harper <james.harper@bendigoit.com.au> wrote:> > > Those DomU->DomU figures (mine and yours) are obviously less than > desirable though, given the DomU->Dom0 performance... I''ll see what I > can do. > > James > >I hope I''m wrong but I''m not sure there''s much to do. I was curious about the numbers people were getting with Windows PV drivers. These match DomU to DomU in Linux PV. Transferring data to a DomU seems to be horribly inefficient and I''m not sure you''ll be able to do anything about it. Seems like it''s a Xen architecture thing. Like I said, I hope I''m wrong. Grant _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> On Mon, Jan 19, 2009 at 3:50 PM, James Harper > <james.harper@bendigoit.com.au> wrote: > > Those DomU->DomU figures (mine and yours) are obviously lessthan> desirable though, given the DomU->Dom0 performance... I''ll seewhat> I can do. > > James > > I hope I''m wrong but I''m not sure there''s much to do. I was curiousabout> the numbers people were getting with Windows PV drivers. These matchDomU> to DomU in Linux PV. Transferring data to a DomU seems to be horribly > inefficient and I''m not sure you''ll be able to do anything about it.Seems> like it''s a Xen architecture thing. Like I said, I hope I''m wrong. >I don''t have a pair of Linux DomU''s to test so I can''t confirm or deny that. Anyone else able to test? James _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
James Harper wrote:>> I hope I''m wrong but I''m not sure there''s much to do. I was curious about >> >> the numbers people were getting with Windows PV drivers. These match DomU >> >> to DomU in Linux PV. > > I don''t have a pair of Linux DomU''s to test so I can''t confirm or deny > that. Anyone else able to test? > >Let''s see. Using HW of the same spec, domU <-> domU (same dom0): - Windows 2003 : 530 Mbps (this is the previous result) - RHEL 5.2 (PV) : 2.7 Gbps I don''t have a pair of Linux HVM domUs with PV drivers to check. I''d say there''s a lot of room for improvement :) Regards, Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> > > Let''s see. Using HW of the same spec, domU <-> domU (same dom0): > - Windows 2003 : 530 Mbps (this is the previous result) > - RHEL 5.2 (PV) : 2.7 Gbps > I don''t have a pair of Linux HVM domUs with PV drivers to check. > > I''d say there''s a lot of room for improvement :) > > Regards, > > Fajar > > _I''m not sure where you''ve gotten those numbers from but I''m curious why I don''t. I have two companies using Xen on CentOS 5.2 with CentOS 5.2 DomUs and I''m getting the same numbers on both. DomU to Dom0 (or anywhere outside) is as fast as the hardware. DomU to DomU runs about 450-500Mbits/sec. It''s consistant so far. If I could get 2.7 Gbps I''d be happy as a clam but I haven''t seen it. Grant McWilliams Some people, when confronted with a problem, think "I know, I''ll use Windows." Now they have two problems. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Grant McWilliams wrote:> > > - Windows 2003 : 530 Mbps (this is the previous result) > - RHEL 5.2 (PV) : 2.7 Gbps > > > > I''m not sure where you''ve gotten those numbers from but I''m curious > why I don''t. I have two companies using Xen on CentOS 5.2 with CentOS > 5.2 DomUsPV, right?> and I''m getting the same numbers on both. DomU to Dom0 (or anywhere > outside) is as fast as the hardware.Which is what? 1 Gbps?> DomU to DomU runs about 450-500Mbits/sec. It''s consistant so far. If I > could get 2.7 Gbps I''d be happy as a clam but I haven''t seen it. >>From what I can tell, network performance in domUs are mostly CPU-bound.I''m using dual X5450@3GHz, disabled hyperthread, dom0 only uses one CPU (CPU0), and both domUs are using different physical CPUs (one for each). Are you perhaps using the same CPU for dom0 and domUs? Or maybe you''re using a newer Xen and enabled cpu_cap? Regards, Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> > James Harper wrote: > >> I hope I''m wrong but I''m not sure there''s much to do. I was curious > about > >> > >> the numbers people were getting with Windows PV drivers. Thesematch> DomU > >> > >> to DomU in Linux PV. > > > > I don''t have a pair of Linux DomU''s to test so I can''t confirm ordeny> > that. Anyone else able to test? > > > > > > Let''s see. Using HW of the same spec, domU <-> domU (same dom0): > - Windows 2003 : 530 Mbps (this is the previous result) > - RHEL 5.2 (PV) : 2.7 Gbps > I don''t have a pair of Linux HVM domUs with PV drivers to check. > > I''d say there''s a lot of room for improvement :) >Sounds like it. James _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hi, Where are the drivers available from? Thanks Ian -----Original Message----- From: xen-users-bounces@lists.xensource.com [mailto:xen-users-bounces@lists.xensource.com] On Behalf Of James Harper Sent: 20 January 2009 09:43 To: Fajar A. Nugraha; xen-users Subject: RE: [Xen-users] GPLPV network performance> > James Harper wrote: > >> I hope I''m wrong but I''m not sure there''s much to do. I was curious > about > >> > >> the numbers people were getting with Windows PV drivers. Thesematch> DomU > >> > >> to DomU in Linux PV. > > > > I don''t have a pair of Linux DomU''s to test so I can''t confirm ordeny> > that. Anyone else able to test? > > > > > > Let''s see. Using HW of the same spec, domU <-> domU (same dom0): > - Windows 2003 : 530 Mbps (this is the previous result) > - RHEL 5.2 (PV) : 2.7 Gbps > I don''t have a pair of Linux HVM domUs with PV drivers to check. > > I''d say there''s a lot of room for improvement :) >Sounds like it. James _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users