Re: [Duplicity-talk] performance issue

[Peter Schuller]

> A while back I noticed a similar issue and wrote a blog post about it:
> http://hewgill.com/journal/entries/366-amazon-s3-performance
> 
> I didn't investigate much further than what's written there, but the
> conclusion is that large files can cause duplicity to slow down a lot,
> and it's probably because of librsync behaviour.

Note that current and past versions of duplicity won't upload
concurrently with local processing. So you should, barring some
indirect effect, not see a decrease in transfer speeds *WHEN a
transfer is in progress*, as a function of the size of the file.

However the volume size in old versions could affect transfer speeds
very negatively, with higher volume sizes. But again independent of
file size.

However, I guess the resolution on the graph is such that you don't
see the variation resulting from individual volumes (especially if
they are 5 MB in size, which is the default). If this is the case,
then as far as I can tell observations would be consistent with
additional bottlenecking on CPU (or disk) locally.

-- 
/ Peter Schuller

PGP userID: 0xE9758B7D or 'Peter Schuller <address@hidden>'
Key retrieval: Send an E-Mail to address@hidden
E-Mail: address@hidden Web: http://www.scode.org

pgpSxyS1OIs8X.pgp
Description: PGP signature