Re: Extraction time of small file from big-ish archive

[Colin Percival <cperciva@tarsnap.com>]

On 10/29/14 11:44, Elvis Stansvik wrote:
> I just signed up for Tarsnap and I'm very pleased with the service. It's a
> weight off my shoulders to finally have both my laptop and my small FreeBSD
> server backed up somewhere. Many thanks to Colin for the nice design of the
> system and the simple tar-like interface.

:-)

> Short question: When extracting a small (few KB) file from a somewhat big
> archive (15 GB or so, with quite a number of files), what is it that takes
> time? For me it took ~4-5 minutes to get such a file out.
> 
> It's not a problem at all since this is something I hope to never have to do,
> but I was just curious what is the actual bottleneck. I'm guessing it's the
> downloading, decryption and scanning of the tar header?

You're exactly right.  Tarsnap stores the tar headers separately from the tar
data in order to speed up this process, but there's still ~50-100 bytes (512
bytes of tar header, compressed) which needs to be downloaded per file so that
tarsnap can decide which parts of the archive need to be extracted.

-- 
Colin Percival
Security Officer Emeritus, FreeBSD | The power to serve
Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid