[gentoo-user] Is a git based tree going to save me bandwidth and time?

[gentoo-user] Is a git based tree going to save me bandwidth and time?

[Sebastián Ramírez Magrí]

Hi folks...

I've been thinking about switching from a rsync based tree to a git
based one cloning [0]. The main reasons because I would do that is in
order to save bandwidth (I've a slow GSM connection in my netbook and
I sync two other gentoo boxes from the first one) and maybe time.

So here goes the question, Is a git based tree really going to save me
an appreciable bandwidth and time on syncing?, Can I keep the same
replication functionality rsync gives me to sync my other boxes?

[0] http://github.com/funtoo/portage/tree/gentoo.org

[gentoo-user] Re: Is a git based tree going to save me bandwidth and time?

[Nikos Chantziaras]

On 08/03/2010 08:03 AM, Sebastián Ramírez Magrí wrote:
>
> Hi folks...
>
> I've been thinking about switching from a rsync based tree to a git
> based one cloning [0]. The main reasons because I would do that is in
> order to save bandwidth (I've a slow GSM connection in my netbook and
> I sync two other gentoo boxes from the first one) and maybe time.
>
> So here goes the question, Is a git based tree really going to save me
> an appreciable bandwidth and time on syncing?, Can I keep the same
> replication functionality rsync gives me to sync my other boxes?
>
> [0] http://github.com/funtoo/portage/tree/gentoo.org

Git needs to move much less data around than rsync.  It only transfers 
differences, not whole files.

Re: [gentoo-user] Re: Is a git based tree going to save me bandwidth and time?

[Graham Murray]

Nikos Chantziaras <realnc@arcor.de> writes:

> Git needs to move much less data around than rsync.  It only transfers
> differences, not whole files.

But is uses a *lot* more disk space on the systems as each system
contains the full history.

Re: [gentoo-user] Re: Is a git based tree going to save me bandwidth and time?

[Michael Schreckenbauer]

Am Dienstag, 3. August 2010, 08:31:38 schrieb Nikos Chantziaras:
> On 08/03/2010 08:03 AM, Sebastián Ramírez Magrí wrote:
> > Hi folks...
> > 
> > I've been thinking about switching from a rsync based tree to a git
> > based one cloning [0]. The main reasons because I would do that is in
> > order to save bandwidth (I've a slow GSM connection in my netbook and
> > I sync two other gentoo boxes from the first one) and maybe time.
> > 
> > So here goes the question, Is a git based tree really going to save me
> > an appreciable bandwidth and time on syncing?, Can I keep the same
> > replication functionality rsync gives me to sync my other boxes?
> > 
> > [0] http://github.com/funtoo/portage/tree/gentoo.org
> 
> Git needs to move much less data around than rsync.  It only transfers
> differences, not whole files.

Not true. rsync uses delta-encoding to minimize data transfers.

Regards
Michael

Re: [gentoo-user] Re: Is a git based tree going to save me bandwidth and time?

[Florian Philipp]

[-- Attachment #1: Type: text/plain, Size: 565 bytes --]

Am 03.08.2010 14:52, schrieb Graham Murray:
> Nikos Chantziaras <realnc@arcor.de> writes:
> 
>> Git needs to move much less data around than rsync.  It only transfers
>> differences, not whole files.
> 
> But is uses a *lot* more disk space on the systems as each system
> contains the full history.
> 

While your statement is correct, you can still avoid having the whole
history and make a shallow clone. It just doesn't help much because you
still need all the metadata:
http://blogs.gnome.org/simos/2009/04/18/git-clones-vs-shallow-git-clones/


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]

Re: [gentoo-user] Re: Is a git based tree going to save me bandwidth and time?

[Florian Philipp]

[-- Attachment #1: Type: text/plain, Size: 665 bytes --]

Am 03.08.2010 16:11, schrieb Michael Schreckenbauer:
> Am Dienstag, 3. August 2010, 08:31:38 schrieb Nikos Chantziaras:
>> On 08/03/2010 08:03 AM, Sebastián Ramírez Magrí wrote:
[...]
>>
>> Git needs to move much less data around than rsync.  It only transfers
>> differences, not whole files.
> 
> Not true. rsync uses delta-encoding to minimize data transfers.
> 

Not necessarily true: Many (all?) public gentoo mirrors deactivate
delta-encoding in order to limit CPU-utilization. I would also guess
that git's delta encoding has a much finer granularity because it works
on lines (in text files) while rsync is designed to work on binary data.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]

Re: [gentoo-user] Is a git based tree going to save me bandwidth and time?

[Sergei Trofimovich]

[-- Attachment #1: Type: text/plain, Size: 546 bytes --]

Hi Sebastián,

> I've been thinking about switching from a rsync based tree to a git
> based one cloning [0]. The main reasons because I would do that is in
> order to save bandwidth (I've a slow GSM connection in my netbook and
> I sync two other gentoo boxes from the first one) and maybe time.

When I had awfully slow internet I used to use app-portage/emerge-delta-webrsync.
emerge-delta-webrsync recreates portage tarball from previous state and patches.
It usually takes about 300KB (one patch size) per day.

-- 

  Sergei

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[gentoo-user] Re: Is a git based tree going to save me bandwidth and time?

[Sebastián Ramírez Magrí]

Sergei Trofimovich <slyfox@gentoo.org> writes:

> Hi Sebastián,
>
>> I've been thinking about switching from a rsync based tree to a git
>> based one cloning [0]. The main reasons because I would do that is in
>> order to save bandwidth (I've a slow GSM connection in my netbook and
>> I sync two other gentoo boxes from the first one) and maybe time.
>
> When I had awfully slow internet I used to use app-portage/emerge-delta-webrsync.
> emerge-delta-webrsync recreates portage tarball from previous state and patches.
> It usually takes about 300KB (one patch size) per day.

I've been using delta-webrsync to update the _main node_ too. I think
git can't really beat delta-webrsync... Will try to do some bandwith
benchmarks and post the results asap...