Hi, I''m investgating network performance recently. Following paper work"Optimizing network performance in Xen ", says network performance in domU suffer about 50% degradation. When I try to rebuild the experiment, it seems not regain the statics that points out significant degradation. (Although two computers used only) Take one as benchmark server running Xen for deifferent configuration (native linux/dom0/domU), the other running in Native Linux is client. The problem of network performance is corrected already? Or I just don''t get the wright experiment to test? _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Fri, Oct 05, 2007 at 03:02:13AM +0000, 阿淳 wrote:> Hi, > I''m investgating network performance recently. > Following paper work"Optimizing network performance in Xen ", says network performance in domU suffer about 50% degradation.Do you have a pointer to this paper? If it''s talking about Xen 3.0.2 or earlier then the problem is already solved as 3.0.3 and later versions use TSO between domU an dom0 to drastically reduce the cost of virtualisation. Cheers, -- Visit Openswan at http://www.openswan.org/ Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
pradeep singh rautela
2007-Oct-09 18:04 UTC
Re: [Xen-devel] Network performance in Xen 3.0.3
On 10/9/07, Herbert Xu <herbert@gondor.apana.org.au> wrote:> On Fri, Oct 05, 2007 at 03:02:13AM +0000, 阿淳 wrote: > > Hi, > > I'm investgating network performance recently. > > Following paper work"Optimizing network performance in Xen ", says network performance in domU suffer about 50% degradation. > > Do you have a pointer to this paper?http://www.cse.psu.edu/~bhuvan/teaching/spring06/papers/xen-net-opt.pdf> If it's talking about Xen 3.0.2 or earlier then the problem is > already solved as 3.0.3 and later versions use TSO between domU > an dom0 to drastically reduce the cost of virtualisation.Herbert, can you let us know how does domU network performance fares with the native stack performance w.r.t PV and HVM guests both. Couldn't find some concrete information on these number sadly. Thanks -- pradeep singh rautela "question = ( to ) ? be : ! be;" -- Wm. Shakespeare _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Tue, Oct 09, 2007 at 11:34:09PM +0530, pradeep singh rautela wrote:> > http://www.cse.psu.edu/~bhuvan/teaching/spring06/papers/xen-net-opt.pdfSeems to be based on Xen 2.0, right? In that case my original comment applies.> > If it''s talking about Xen 3.0.2 or earlier then the problem is > > already solved as 3.0.3 and later versions use TSO between domU > > an dom0 to drastically reduce the cost of virtualisation.> Herbert, can you let us know how does domU network performance fares > with the native stack performance w.r.t PV and HVM guests both. > Couldn''t find some concrete information on these number sadly.Pure HVM drivers are still as bad as they ever were. However, PV-on-HVM drivers lets you get similar performance to PV drivers in an HVM domain. Cheers, -- Visit Openswan at http://www.openswan.org/ Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
pradeep singh rautela
2007-Oct-10 03:31 UTC
Re: [Xen-devel] Network performance in Xen 3.0.3
On 10/10/07, Herbert Xu <herbert@gondor.apana.org.au> wrote:> On Tue, Oct 09, 2007 at 11:34:09PM +0530, pradeep singh rautela wrote: > > > > http://www.cse.psu.edu/~bhuvan/teaching/spring06/papers/xen-net-opt.pdf > > Seems to be based on Xen 2.0, right?Yes, i missed that part earlier i guess.> In that case my original comment applies. > > > > If it''s talking about Xen 3.0.2 or earlier then the problem is > > > already solved as 3.0.3 and later versions use TSO between domU > > > an dom0 to drastically reduce the cost of virtualisation. > > > Herbert, can you let us know how does domU network performance fares > > with the native stack performance w.r.t PV and HVM guests both. > > Couldn''t find some concrete information on these number sadly. > > Pure HVM drivers are still as bad as they ever were. However, > PV-on-HVM drivers lets you get similar performance to PV drivers > in an HVM domain.Thanks for the info Herbet. Cheers, -- pradeep singh rautela "question = ( to ) ? be : ! be;" -- Wm. Shakespeare _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel