You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Bob Archer <Bo...@amsi.com> on 2008/12/10 21:02:27 UTC
svn vs http performance
Hello all,
I was wondering if there is much of a difference between svn: and http:
performance. I have read that svn is faster but is it "a lot" faster or
just a bit?
I basically want to move from a password file to using NT Auth/Active
Directory so I don't have to manage username/passwords and which cause
me to just give people simple passwords that never change. It seems the
easiest/best way to do this is to go to Apache rather than svnserve. I
haven't found a way to use AD with svnserve for Tortoise client... if
there is an easy way to do this without building client exe's please
enlighten me.
BTW: This is all inside the firewall stuff. With remote offices on the
WAN being some type of tunneled IP/VPN stuff over the internet.
Thanks,
BOb
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=982542
To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].
Re: svn vs http performance
Posted by Ivan Zhakov <iv...@visualsvn.com>.
On Thu, Dec 11, 2008 at 12:02 AM, Bob Archer <Bo...@amsi.com> wrote:
> Hello all,
>
>
>
> I was wondering if there is much of a difference between svn: and http:
> performance. I have read that svn is faster but is it "a lot" faster or just
> a bit?
>
Hi Bob,
Currently http: is significantly slower over WAN networks, but you
should not see big difference over LAN networks. But performance of
http protocol should be improved in next versions of Subversion.
--
Ivan Zhakov
VisualSVN Team
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=982878
To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].
RE: svn vs http performance
Posted by Bob Archer <Bo...@amsi.com>.
Tony,
Thanks for the info. We are running svn on a Windows server, so I think
that precludes using ssh.
I'm not sure what auth method I will be using. I was planning on using
the visualSVN server and I don't know exactly which authentication
module it uses. I really don't want a 50% hit as another poster here
mentioned that they are seeing.
BOb
-----Original Message-----
From: Tony Butt [mailto:tjb@cea.com.au]
Sent: Wednesday, December 10, 2008 8:43 PM
To: users@subversion.tigris.org
Subject: Re: svn vs http performance
On Wed, 2008-12-10 at 16:02 -0500, Bob Archer wrote:
> Hello all,
>
>
>
> I was wondering if there is much of a difference between svn: and
> http: performance. I have read that svn is faster but is it "a lot"
> faster or just a bit?
>
That depends on how your authentication is done.
We used an Active Directory server for authentication, and ran
subversion with mod_auth_krb on a Suse Linux Enterprise Server and
apache 2.0, and found that svn+ssh:// was much faster than http://
That was because there was no caching of authentication credentials
between GET's for files, so every file fetched in a checkout (for
instance) would cause the kerberos credentials to be checked again.
We eventually switched to mod_auth_pam, and used kerberos in PAM to
authenticate, which performed better, and gave us single sign on. (svn
+ssh uses PAM too). The svn+ssh:// was still at least 2x faster than
http://
We currently use mod_authnz_ldap with an apache 2.2 server, and http://
is now pretty good, as there is authentication caching happening on
apache. I still think svn+ssh is a little faster, but only by a small
margin.
Tony Butt
Senior Software Engineer
CEA Technologies
Canberra
Australia
>
>
> I basically want to move from a password file to using NT Auth/Active
> Directory so I don't have to manage username/passwords and which cause
> me to just give people simple passwords that never change. It seems
> the easiest/best way to do this is to go to Apache rather than
> svnserve. I haven't found a way to use AD with svnserve for Tortoise
> client... if there is an easy way to do this without building client
> exe's please enlighten me.
>
>
>
> BTW: This is all inside the firewall stuff. With remote offices on the
> WAN being some type of tunneled IP/VPN stuff over the internet.
>
>
>
> Thanks,
>
> BOb
>
>
>
>
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageI
d=982636
To unsubscribe from this discussion, e-mail:
[users-unsubscribe@subversion.tigris.org].
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=982870
To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].
Re: svn vs http performance
Posted by Steve Bakke <st...@amd.com>.
On Dec 10, 2008, at 8:43 PM, Tony Butt wrote:
> On Wed, 2008-12-10 at 16:02 -0500, Bob Archer wrote:
>> Hello all,
>>
>>
>>
>> I was wondering if there is much of a difference between svn: and
>> http: performance. I have read that svn is faster but is it “a lot”
>> faster or just a bit?
>>
> That depends on how your authentication is done.
> We used an Active Directory server for authentication, and ran
> subversion with mod_auth_krb on a Suse Linux Enterprise Server and
> apache 2.0, and found that svn+ssh:// was much faster than http://
>
> That was because there was no caching of authentication credentials
> between GET's for files, so every file fetched in a checkout (for
> instance) would cause the kerberos credentials to be checked again.
>
> We eventually switched to mod_auth_pam, and used kerberos in PAM to
> authenticate, which performed better, and gave us single sign on. (svn
> +ssh uses PAM too). The svn+ssh:// was still at least 2x faster than
> http://
>
In our testing, we found that svn: was about 2x faster than svn+ssh:
for the way we tend to access our repository.
-steve
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=982669
To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].
Re: svn vs http performance
Posted by Tony Butt <tj...@cea.com.au>.
On Wed, 2008-12-10 at 16:02 -0500, Bob Archer wrote:
> Hello all,
>
>
>
> I was wondering if there is much of a difference between svn: and
> http: performance. I have read that svn is faster but is it “a lot”
> faster or just a bit?
>
That depends on how your authentication is done.
We used an Active Directory server for authentication, and ran
subversion with mod_auth_krb on a Suse Linux Enterprise Server and
apache 2.0, and found that svn+ssh:// was much faster than http://
That was because there was no caching of authentication credentials
between GET's for files, so every file fetched in a checkout (for
instance) would cause the kerberos credentials to be checked again.
We eventually switched to mod_auth_pam, and used kerberos in PAM to
authenticate, which performed better, and gave us single sign on. (svn
+ssh uses PAM too). The svn+ssh:// was still at least 2x faster than
http://
We currently use mod_authnz_ldap with an apache 2.2 server, and http://
is now pretty good, as there is authentication caching happening on
apache. I still think svn+ssh is a little faster, but only by a small
margin.
Tony Butt
Senior Software Engineer
CEA Technologies
Canberra
Australia
>
>
> I basically want to move from a password file to using NT Auth/Active
> Directory so I don’t have to manage username/passwords and which cause
> me to just give people simple passwords that never change. It seems
> the easiest/best way to do this is to go to Apache rather than
> svnserve. I haven’t found a way to use AD with svnserve for Tortoise
> client… if there is an easy way to do this without building client
> exe’s please enlighten me.
>
>
>
> BTW: This is all inside the firewall stuff. With remote offices on the
> WAN being some type of tunneled IP/VPN stuff over the internet.
>
>
>
> Thanks,
>
> BOb
>
>
>
>
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=982636
To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].