OpenVZ Forum


Home » Mailing lists » Users » Re: [Announce] Kernel RHEL6 testing 042stab053.4
Re: [Announce] Kernel RHEL6 testing 042stab053.4 [message #45649] Tue, 27 March 2012 00:01 Go to next message
jjs - mainphrame is currently offline  jjs - mainphrame
Messages: 44
Registered: January 2012
Member
I'm happy to report that the ploop-based CT under 042stab053.4 finishes a
punishing dbench run with no problems. I was curious to see what
performance would be like, and the dbench results from the ploop-based CT
are actually closer to the host performance than the simfs-based CT is.

Joe

On Mon, Mar 26, 2012 at 7:05 AM, Kir Kolyshkin <kir@openvz.org> wrote:

> OpenVZ project has released a new RHEL6 based testing kernel. Read below
> for more information. Everyone using this kernel branch is advised to
> upgrade.
>
> NOTE this is a *testing* kernel, not recommended for production.
>
>
> Changes
> =======
> (since 042stab053.3)
> * Fixes in UBC, CPT, ploop
>
>
> Compatibility
> =============
> No new issues
>
>
> Download
> ========
> http://wiki.openvz.org/**Download/kernel/rhel6-testing/**042 stab053.4< http://wiki.openvz.org/Download/kernel/rhel6-testing/042stab 053.4>
>
>
> Bug reporting
> =============
> Use http://bugzilla.openvz.org/ to report any bugs found.
>
>
> Other sources of info on updates
> ==============================**==
> See http://wiki.openvz.org/News to view all the news (including updates)
> online. There you can also find RSS/Atom feed links.
>
>
> Best regards,
> OpenVZ team.
> ______________________________**_________________
> Announce mailing list
> Announce@openvz.org
> https://openvz.org/mailman/**listinfo/announce<https://openvz.org/mailman/listinfo/announce>
>


http://static.openvz.org/userbars/openvz-user.png
Re: Re: [Announce] Kernel RHEL6 testing 042stab053.4 [message #45650 is a reply to message #45649] Tue, 27 March 2012 04:45 Go to previous messageGo to next message
Kirill Korotaev is currently offline  Kirill Korotaev
Messages: 137
Registered: January 2006
Senior Member
:)))
Actually, this can't be true, cause simfs overhead is naturally 0%. It intercepts only statfs() to fix df output, all the rest calls are native. So it's statistical deviation if you see simfs to be slower then native.

And, yes, ploop was designed to have near 0% overhead as well (unlike Linux loopback). It bypasses FS layer and works directly with block layer on fast paths.

Thanks,
Kirill


On Mar 27, 2012, at 04:01 , jjs - mainphrame wrote:

> I'm happy to report that the ploop-based CT under 042stab053.4 finishes a punishing dbench run with no problems. I was curious to see what performance would be like, and the dbench results from the ploop-based CT are actually closer to the host performance than the simfs-based CT is.
>
> Joe
>
> On Mon, Mar 26, 2012 at 7:05 AM, Kir Kolyshkin <kir@openvz.org> wrote:
> OpenVZ project has released a new RHEL6 based testing kernel. Read below
> for more information. Everyone using this kernel branch is advised to
> upgrade.
>
> NOTE this is a *testing* kernel, not recommended for production.
>
>
> Changes
> =======
> (since 042stab053.3)
> * Fixes in UBC, CPT, ploop
>
>
> Compatibility
> =============
> No new issues
>
>
> Download
> ========
> http://wiki.openvz.org/Download/kernel/rhel6-testing/042stab 053.4
>
>
> Bug reporting
> =============
> Use http://bugzilla.openvz.org/ to report any bugs found.
>
>
> Other sources of info on updates
> ================================
> See http://wiki.openvz.org/News to view all the news (including updates)
> online. There you can also find RSS/Atom feed links.
>
>
> Best regards,
> OpenVZ team.
> _______________________________________________
> Announce mailing list
> Announce@openvz.org
> https://openvz.org/mailman/listinfo/announce
>
> <ATT00001.c>
Re: Re: [Announce] Kernel RHEL6 testing 042stab053.4 [message #45651 is a reply to message #45650] Tue, 27 March 2012 05:29 Go to previous messageGo to next message
jjs - mainphrame is currently offline  jjs - mainphrame
Messages: 44
Registered: January 2012
Member
Kir,

I'll send you the dbench scripts and output if you want - I'll even give
you remote access to my test system if you want to take a look.

Joe

On Mon, Mar 26, 2012 at 9:45 PM, Kirill Korotaev <dev@parallels.com> wrote:

> :)))
> Actually, this can't be true, cause simfs overhead is naturally 0%. It
> intercepts only statfs() to fix df output, all the rest calls are native.
> So it's statistical deviation if you see simfs to be slower then native.
>
> And, yes, ploop was designed to have near 0% overhead as well (unlike
> Linux loopback). It bypasses FS layer and works directly with block layer
> on fast paths.
>
> Thanks,
> Kirill
>
>
> On Mar 27, 2012, at 04:01 , jjs - mainphrame wrote:
>
> > I'm happy to report that the ploop-based CT under 042stab053.4 finishes
> a punishing dbench run with no problems. I was curious to see what
> performance would be like, and the dbench results from the ploop-based CT
> are actually closer to the host performance than the simfs-based CT is.
> >
> > Joe
> >
> > On Mon, Mar 26, 2012 at 7:05 AM, Kir Kolyshkin <kir@openvz.org> wrote:
> > OpenVZ project has released a new RHEL6 based testing kernel. Read below
> > for more information. Everyone using this kernel branch is advised to
> > upgrade.
> >
> > NOTE this is a *testing* kernel, not recommended for production.
> >
> >
> > Changes
> > =======
> > (since 042stab053.3)
> > * Fixes in UBC, CPT, ploop
> >
> >
> > Compatibility
> > =============
> > No new issues
> >
> >
> > Download
> > ========
> > http://wiki.openvz.org/Download/kernel/rhel6-testing/042stab 053.4
> >
> >
> > Bug reporting
> > =============
> > Use http://bugzilla.openvz.org/ to report any bugs found.
> >
> >
> > Other sources of info on updates
> > ================================
> > See http://wiki.openvz.org/News to view all the news (including updates)
> > online. There you can also find RSS/Atom feed links.
> >
> >
> > Best regards,
> > OpenVZ team.
> > _______________________________________________
> > Announce mailing list
> > Announce@openvz.org
> > https://openvz.org/mailman/listinfo/announce
> >
> > <ATT00001.c>
>
>


http://static.openvz.org/userbars/openvz-user.png
Re: Re: [Announce] Kernel RHEL6 testing 042stab053.4 [message #45659 is a reply to message #45650] Tue, 27 March 2012 20:34 Go to previous message
jjs - mainphrame is currently offline  jjs - mainphrame
Messages: 44
Registered: January 2012
Member
Kir, you may be right. A dbench run with ploop CT yesterday yielded the
highest performance numbers ever seen on that hardware, but subsequent runs
have not been able to reproduce those results, except on the host itself.

Joe

On Mon, Mar 26, 2012 at 9:45 PM, Kirill Korotaev <dev@parallels.com> wrote:

> :)))
> Actually, this can't be true, cause simfs overhead is naturally 0%. It
> intercepts only statfs() to fix df output, all the rest calls are native.
> So it's statistical deviation if you see simfs to be slower then native.
>
> And, yes, ploop was designed to have near 0% overhead as well (unlike
> Linux loopback). It bypasses FS layer and works directly with block layer
> on fast paths.
>
> Thanks,
> Kirill
>
>
> On Mar 27, 2012, at 04:01 , jjs - mainphrame wrote:
>
> > I'm happy to report that the ploop-based CT under 042stab053.4 finishes
> a punishing dbench run with no problems. I was curious to see what
> performance would be like, and the dbench results from the ploop-based CT
> are actually closer to the host performance than the simfs-based CT is.
> >
> > Joe
> >
> > On Mon, Mar 26, 2012 at 7:05 AM, Kir Kolyshkin <kir@openvz.org> wrote:
> > OpenVZ project has released a new RHEL6 based testing kernel. Read below
> > for more information. Everyone using this kernel branch is advised to
> > upgrade.
> >
> > NOTE this is a *testing* kernel, not recommended for production.
> >
> >
> > Changes
> > =======
> > (since 042stab053.3)
> > * Fixes in UBC, CPT, ploop
> >
> >
> > Compatibility
> > =============
> > No new issues
> >
> >
> > Download
> > ========
> > http://wiki.openvz.org/Download/kernel/rhel6-testing/042stab 053.4
> >
> >
> > Bug reporting
> > =============
> > Use http://bugzilla.openvz.org/ to report any bugs found.
> >
> >
> > Other sources of info on updates
> > ================================
> > See http://wiki.openvz.org/News to view all the news (including updates)
> > online. There you can also find RSS/Atom feed links.
> >
> >
> > Best regards,
> > OpenVZ team.
> > _______________________________________________
> > Announce mailing list
> > Announce@openvz.org
> > https://openvz.org/mailman/listinfo/announce
> >
> > <ATT00001.c>
>
>


http://static.openvz.org/userbars/openvz-user.png
Previous Topic: Problems encountered increasing CT diskspace with layout=ploop
Next Topic: ploop
Goto Forum:
  


Current Time: Tue Dec 03 09:05:47 GMT 2024

Total time taken to generate the page: 0.17691 seconds