You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Leszek Szarlej <le...@gmail.com> on 2010/03/09 03:03:33 UTC

svn 1.6x and apache compression - mod_deflate

Hi

In svn book ver. 1.0 there is written:
"Deflate compression places a small burden on the client and server to
compress and decompress network transmissions as a way to minimize the size
of the actual transmission. In cases where network bandwidth is in short
supply, this kind of compression can greatly increase the speed at which
communications between server and client can be sent. In extreme cases, this
minimized network transmission could be the difference between an operation
timing out or completing successfully."
In the same book in version 1.5 there is no word about why/why not. Is it
mean that mod_deflate is not recomended for current versions of svn? I did
tests on svn 1.6.9 and compression rate is about 30-40 percent and checkout
times are ~30%  less

Thanks
Leszek

Re: svn 1.6x and apache compression - mod_deflate

Posted by Leszek Szarlej <le...@gmail.com>.
That would be strange for me as far as it looks for Accept-Encoding: gzip
header

 I've checked current day:
bash-3.2$ cat access_log.2010-03-10 |grep TortoiseSVN-1.5 |wc -l
1077

I revieved some of them and most of them have resp code
200 or 207
207 Multi-Status (WebDAV) (RFC 4918)

There is also variable to put in custom log for logging if client accepts
compression.
I think I will test it.

Regards
Leszek

On 10 March 2010 18:31, Mark Phippard <ma...@gmail.com> wrote:

> Do a checkout or command using a SVN client that does not have deflate
> support.  For example, the TortoiseSVN 1.5.x clients.  As soon as you
> do this, your server will probably crash.
>
> When the SVN clients have deflate support it all works great.  The bug
> happens when a client without support hits the server.  This could
> also include command line tools like wget.
>
> Mark
>
>
> On Wed, Mar 10, 2010 at 11:58 AM, Leszek Szarlej
> <le...@gmail.com> wrote:
> > Hi I am sharing with you my results of running with mod_deflate and svn
> > 1.9.9 for ~13 hours. I hope attachments works here.. see attached file
> >
> > The load of the server decreased but there were also other changes
> applied
> > in the same time like LDAP cache or ServerTokens PROD
> >
> > To this time mod_deflate works very efficiently without notificable
> > burden of server.
> >
> >
> >
> >
> >
> >
> >
> > Regards
> > Leszek
> >
> > On 9 March 2010 17:33, Stefan Sperling <st...@elego.de> wrote:
> >>
> >> On Tue, Mar 09, 2010 at 05:10:11PM +0100, Leszek Szarlej wrote:
> >> > Thanks for the info.
> >> > >See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
> >> > In mentioned example apache runs on worker mpm. I've sent a message to
> >> > this
> >> > group few days ago about users experience with svn and worker mpm. I
> >> > didn't
> >> > get any answer.
> >>
> >> Can you provide a link to this message so it's easy to find it and to
> >> reply to it?
> >> The archives are at
> >> http://mail-archives.apache.org/mod_mbox/subversion-dev/
> >>
> >> > It concerns me if svn is thread-safe and if the said memory
> >> > problem would appear on prefork mpm. I will be doing some tests this
> >> > week.
> >>
> >> Thanks, tests always help and give insight.
> >>
> >> Note that the problem is not specific to svn. If I understand correctly
> >> it's a problem with mod_deflate. It's possible to work around it in
> >> Subversion, but a proper fix would have to be made within mod_deflate.
> >>
> >> I don't know what the status of this bug is with the httpd team,
> >> whether it's known there, or if a fix is in sight.
> >> The fix seems to be important to you, so maybe you could ask them?
> >>
> >> Thanks,
> >> Stefan
> >
> >
>
>
>
> --
> Thanks
>
> Mark Phippard
> http://markphip.blogspot.com/
>

Re: svn 1.6x and apache compression - mod_deflate

Posted by Mark Phippard <ma...@gmail.com>.
Do a checkout or command using a SVN client that does not have deflate
support.  For example, the TortoiseSVN 1.5.x clients.  As soon as you
do this, your server will probably crash.

When the SVN clients have deflate support it all works great.  The bug
happens when a client without support hits the server.  This could
also include command line tools like wget.

Mark


On Wed, Mar 10, 2010 at 11:58 AM, Leszek Szarlej
<le...@gmail.com> wrote:
> Hi I am sharing with you my results of running with mod_deflate and svn
> 1.9.9 for ~13 hours. I hope attachments works here.. see attached file
>
> The load of the server decreased but there were also other changes applied
> in the same time like LDAP cache or ServerTokens PROD
>
> To this time mod_deflate works very efficiently without notificable
> burden of server.
>
>
>
>
>
>
>
> Regards
> Leszek
>
> On 9 March 2010 17:33, Stefan Sperling <st...@elego.de> wrote:
>>
>> On Tue, Mar 09, 2010 at 05:10:11PM +0100, Leszek Szarlej wrote:
>> > Thanks for the info.
>> > >See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
>> > In mentioned example apache runs on worker mpm. I've sent a message to
>> > this
>> > group few days ago about users experience with svn and worker mpm. I
>> > didn't
>> > get any answer.
>>
>> Can you provide a link to this message so it's easy to find it and to
>> reply to it?
>> The archives are at
>> http://mail-archives.apache.org/mod_mbox/subversion-dev/
>>
>> > It concerns me if svn is thread-safe and if the said memory
>> > problem would appear on prefork mpm. I will be doing some tests this
>> > week.
>>
>> Thanks, tests always help and give insight.
>>
>> Note that the problem is not specific to svn. If I understand correctly
>> it's a problem with mod_deflate. It's possible to work around it in
>> Subversion, but a proper fix would have to be made within mod_deflate.
>>
>> I don't know what the status of this bug is with the httpd team,
>> whether it's known there, or if a fix is in sight.
>> The fix seems to be important to you, so maybe you could ask them?
>>
>> Thanks,
>> Stefan
>
>



-- 
Thanks

Mark Phippard
http://markphip.blogspot.com/

Re: svn 1.6x and apache compression - mod_deflate

Posted by Leszek Szarlej <le...@gmail.com>.
Hi I am sharing with you my results of running with mod_deflate and svn
1.9.9 for ~13 hours. I hope attachments works here.. see attached file

The load of the server decreased but there were also other changes applied
in the same time like LDAP cache or ServerTokens PROD

To this time mod_deflate works very efficiently without notificable
burden of server.






Regards
Leszek

On 9 March 2010 17:33, Stefan Sperling <st...@elego.de> wrote:

> On Tue, Mar 09, 2010 at 05:10:11PM +0100, Leszek Szarlej wrote:
> > Thanks for the info.
> > >See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
> > In mentioned example apache runs on worker mpm. I've sent a message to
> this
> > group few days ago about users experience with svn and worker mpm. I
> didn't
> > get any answer.
>
> Can you provide a link to this message so it's easy to find it and to
> reply to it?
> The archives are at
> http://mail-archives.apache.org/mod_mbox/subversion-dev/
>
> > It concerns me if svn is thread-safe and if the said memory
> > problem would appear on prefork mpm. I will be doing some tests this
> week.
>
> Thanks, tests always help and give insight.
>
> Note that the problem is not specific to svn. If I understand correctly
> it's a problem with mod_deflate. It's possible to work around it in
> Subversion, but a proper fix would have to be made within mod_deflate.
>
> I don't know what the status of this bug is with the httpd team,
> whether it's known there, or if a fix is in sight.
> The fix seems to be important to you, so maybe you could ask them?
>
> Thanks,
> Stefan
>

Re: svn 1.6x and apache compression - mod_deflate

Posted by Leszek Szarlej <le...@gmail.com>.
Hi I am sharing with you my results of running with mod_deflate and svn
1.9.9 for ~13 hours. I hope attachments works here.. see attached file

The load of the server decreased but there were also other changes applied
in the same time like LDAP cache or ServerTokens PROD

To this time mod_deflate works very efficiently without notificable
burden of server.






Regards
Leszek

On 9 March 2010 17:33, Stefan Sperling <st...@elego.de> wrote:

> On Tue, Mar 09, 2010 at 05:10:11PM +0100, Leszek Szarlej wrote:
> > Thanks for the info.
> > >See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
> > In mentioned example apache runs on worker mpm. I've sent a message to
> this
> > group few days ago about users experience with svn and worker mpm. I
> didn't
> > get any answer.
>
> Can you provide a link to this message so it's easy to find it and to
> reply to it?
> The archives are at
> http://mail-archives.apache.org/mod_mbox/subversion-dev/
>
> > It concerns me if svn is thread-safe and if the said memory
> > problem would appear on prefork mpm. I will be doing some tests this
> week.
>
> Thanks, tests always help and give insight.
>
> Note that the problem is not specific to svn. If I understand correctly
> it's a problem with mod_deflate. It's possible to work around it in
> Subversion, but a proper fix would have to be made within mod_deflate.
>
> I don't know what the status of this bug is with the httpd team,
> whether it's known there, or if a fix is in sight.
> The fix seems to be important to you, so maybe you could ask them?
>
> Thanks,
> Stefan
>

Re: svn 1.6x and apache compression - mod_deflate

Posted by Stefan Sperling <st...@elego.de>.
On Tue, Mar 09, 2010 at 05:33:34PM +0100, Stefan Sperling wrote:
> On Tue, Mar 09, 2010 at 05:10:11PM +0100, Leszek Szarlej wrote:
> > Thanks for the info.
> > >See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
> > In mentioned example apache runs on worker mpm. I've sent a message to this
> > group few days ago about users experience with svn and worker mpm. I didn't
> > get any answer.
> 
> Can you provide a link to this message so it's easy to find it and to
> reply to it?
> The archives are at http://mail-archives.apache.org/mod_mbox/subversion-dev/

Sorry, I meant http://mail-archives.apache.org/mod_mbox/subversion-users/

Stefan

Re: svn 1.6x and apache compression - mod_deflate

Posted by Stefan Sperling <st...@elego.de>.
On Tue, Mar 09, 2010 at 05:10:11PM +0100, Leszek Szarlej wrote:
> Thanks for the info.
> >See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
> In mentioned example apache runs on worker mpm. I've sent a message to this
> group few days ago about users experience with svn and worker mpm. I didn't
> get any answer.

Can you provide a link to this message so it's easy to find it and to
reply to it?
The archives are at http://mail-archives.apache.org/mod_mbox/subversion-dev/

> It concerns me if svn is thread-safe and if the said memory
> problem would appear on prefork mpm. I will be doing some tests this week.

Thanks, tests always help and give insight.

Note that the problem is not specific to svn. If I understand correctly
it's a problem with mod_deflate. It's possible to work around it in
Subversion, but a proper fix would have to be made within mod_deflate.

I don't know what the status of this bug is with the httpd team,
whether it's known there, or if a fix is in sight.
The fix seems to be important to you, so maybe you could ask them?

Thanks,
Stefan

Re: svn 1.6x and apache compression - mod_deflate

Posted by Leszek Szarlej <le...@gmail.com>.
Thanks for the info.
>See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
In mentioned example apache runs on worker mpm. I've sent a message to this
group few days ago about users experience with svn and worker mpm. I didn't
get any answer. It concerns me if svn is thread-safe and if the said memory
problem would appear on prefork mpm. I will be doing some tests this week.
Regards
Leszek

On 9 March 2010 11:05, Stefan Sperling <st...@elego.de> wrote:

> On Tue, Mar 09, 2010 at 04:03:33AM +0100, Leszek Szarlej wrote:
> > Hi
> >
> > In svn book ver. 1.0 there is written:
> > "Deflate compression places a small burden on the client and server to
> > compress and decompress network transmissions as a way to minimize the
> size
> > of the actual transmission. In cases where network bandwidth is in short
> > supply, this kind of compression can greatly increase the speed at which
> > communications between server and client can be sent. In extreme cases,
> this
> > minimized network transmission could be the difference between an
> operation
> > timing out or completing successfully."
> > In the same book in version 1.5 there is no word about why/why not. Is it
> > mean that mod_deflate is not recomended for current versions of svn? I
> did
> > tests on svn 1.6.9 and compression rate is about 30-40 percent and
> checkout
> > times are ~30%  less
>
> I don't know why that paragraph was dropped from the book.
>
> But there's an unrelated problem with using mod_deflate together with
> mod_dav_svn. It can cause memory leaks.
> See http://svn.haxx.se/dev/archive-2009-08/0274.shtml
>
> Stefan
>

Re: svn 1.6x and apache compression - mod_deflate

Posted by Stefan Sperling <st...@elego.de>.
On Tue, Mar 09, 2010 at 04:03:33AM +0100, Leszek Szarlej wrote:
> Hi
> 
> In svn book ver. 1.0 there is written:
> "Deflate compression places a small burden on the client and server to
> compress and decompress network transmissions as a way to minimize the size
> of the actual transmission. In cases where network bandwidth is in short
> supply, this kind of compression can greatly increase the speed at which
> communications between server and client can be sent. In extreme cases, this
> minimized network transmission could be the difference between an operation
> timing out or completing successfully."
> In the same book in version 1.5 there is no word about why/why not. Is it
> mean that mod_deflate is not recomended for current versions of svn? I did
> tests on svn 1.6.9 and compression rate is about 30-40 percent and checkout
> times are ~30%  less

I don't know why that paragraph was dropped from the book.

But there's an unrelated problem with using mod_deflate together with
mod_dav_svn. It can cause memory leaks.
See http://svn.haxx.se/dev/archive-2009-08/0274.shtml

Stefan