Re: bandwidth and space required to sync a basic mirror

anarcat <[email protected]>
Newsgroups gmane.linux.debian.user.mirrors
Message-ID <[email protected]>
On 2013-11-20 13:56:09, Carsten Otto wrote:
> ftp2.de.debian.org reporting in :)

Hi!

Sorry for the immensely late response, but I got caught up on other
things. :)

> On Wed, Nov 20, 2013 at 01:31:00PM -0500, anarcat wrote:
>> Munin/Cacti graph of the disk usage from an existing mirror for the last
>> year would be awesome here, for example.
>
> Debian

[...]

That's exactly what I was looking for, thanks!!!

Out of curiosity, which hardware are you using to host this? a dedicated
server? something with a SAN behind?

>> Bandwidth for updates
>> =====================
>> 
>> It is unclear how much data is transfered on a daily or (ideally)
>> monthly basis for updates to the archive. I found a graph[2] which seem
>> to show the size transfered varies between 2 and 15 GB on a daily basis,
>> but that data is old (2012-08-12) and it's hard to get a good idea of
>> what it actually means - what's the average, for example? Or better,
>> what's the 95th percentile?
>
> The red lines (rsync) correspond to downloads we actively do using rsync
> (after being triggered).
>
> http://ftp.halifax.rwth-aachen.de/~cotto/rrd/debian-backports%23month.png
> http://ftp.halifax.rwth-aachen.de/~cotto/rrd/debian-cd%23month.png
> http://ftp.halifax.rwth-aachen.de/~cotto/rrd/debian-volatile%23month.png
> http://ftp.halifax.rwth-aachen.de/~cotto/rrd/debian%23month.png
>
> The graphical representation is not too helpful, but I also have the
> corresponding raw (Java-)RRD data. Just message me if you need something.

It would be great to have a 95th percentile of that red line, actually.

One of the things I am wondering is if it is worthwhile to keep a
private mirror running for clusters that have lots of machines and want
to spare the bandwidth costs yet do not want to offer the public hosting
commitment...

>> Bandwidth for users
>> ===================
>> 
>> I assume this is the hardest part to figure out, as this will vary
>> wildly according to the region you're in and the relative popularity of
>> your mirror.
>> 
>> The official mirror page[3] states that you need a "T1 or better" (for
>> the record and those who weren't around in those days, a T1 is
>> 1.5Mbit/s). I hardly think this requirement is fit for a public mirror
>> these days. :) The official page does mention this requirement is
>> "higher" for "developped countries", but fails to clarify that
>> requirement.
>
> Have a look at the graphs above. With 51 MBit/sec on average, Debian is
> not very demanding (although we're not ftp.de.debian.org, just ftp2).

I would consider 50mbps a respectable usage. :) I'm curious to see what
kind of pain ftp.de is suffering... ;)

Thank you so much for this again (and everyone else that answered!),
it's great to see those graphs.

A.

-- 
L'art n'est pas un bureau d'anthropométrie.
                        - Léo Ferré, "Préface"
signature.asc (application/pgp-signature, 818 B)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
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=ZFd0
-----END PGP SIGNATURE-----
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.