bzr speed

[John A Meinel]

James Henstridge wrote:
> On 08/02/06, Diwaker Gupta <diwaker.lists at gmail.com> wrote:
>> On 2/7/06, Robert Collins <robertc at robertcollins.net> wrote:
>>> On Tue, 2006-02-07 at 18:58 -0800, Diwaker Gupta wrote:
>>>> Two days back I did a 'bzr branch' on the main bzr.dev branch. I was
>>>> on a reasonably fast cable line, and it took *more than one hour* for
>>>> the operation to finish.
>>> What bzr were you using ? If you where using 0.7 its largely fixed in
>>> bzr.dev itself.
>> I was indeed using bzr 0.7. After my pull yesterday, I installed bzr
>> from bzr.dev and 'bzr version' now reports 0.8pre. So I thought I'll
>> do a clean pull again today -- it still took MORE than an hour.
> 
> The initial pull downloads the entire history, which is ~ 34MB right now.
> 
> There is definitely room to improve the HTTP transport though, including:
>  * support keep alive requests, to reduce TCP connection setup overhead
>  * support compressed response bodies.
> 
> The second one here would require the user to configure their web
> server to support compression too, but should improve things when it
> is supported.
> 
> James.
> 
> 

I actually wrote code to support having compressed weaves, rather than
compressing them on the fly. This was very useful all around, as it also
decreased the local footprint. It did have a slight speed penalty for
local branches. But it was moderate, especially considering we were
comparing with extra large trees.

I think the biggest reason I didn't push for it to land, was I was
hoping that knits (which add compression, along with lots of other
performance benefits) would land, and make my work obsolete.

John
=:->


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060208/864e242a/attachment.pgp