yum-fastestmirror is not the default, it should be

[Fernando Cassia]

On Sun, May 5, 2013 at 2:25 PM, Chris Adams <cmadams at hiwaay.net> wrote:

>
> Yes.  It assumes that low latency equals high throughput, which is not
> always the case (in fact it is often not true).  Now, mirrormanager
> should be pointing you to a regional mirror to start with, so you should
> probably check that (there could be a bad mirror listed or something
> like that).


Well, down here in .ar you get MUCH better bandwidth (usually what you pay
for) for LOCAL connections (*.ar) while you get, if you're lucky 60% of the
paid bandwidth for international connections (with some exceptions, ie
Google, Youtube, which have local servers-caching of some sort), due to all
the traffic shaping and bandwidth limit rules applied to international
connections.

For me, yum-fastestmirror works WONDERFUL.

Here's the output of installing gparted in 12 secs

http://pastebin.com/raw.php?i=Tv8CVHgX

vs 34 seconds without yum-fastestmirror

http://pastebin.com/raw.php?i=rBkVpCDv

Of course for a small package like gparted it's no big deal, but you see
the speeds are 3x better with yum-fastestmirror, and it suddenly turns into
a lot of time when installing big packages...

So, I don't know about what approach it uses, but it works for me.

If you think the current measurements by yum-fastestmirror is broken in
some instances, it's an opportunity to improve the algorithms (ie measuring
the speed of retrieving updates/primary_db the first time it is run, from
the nearest servers, first by domain name / country code and then
geographically? (ie neighboring countries)

FC
-- 
During times of Universal Deceit, telling the truth becomes a revolutionary
act
- George Orwell
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.fedoraproject.org/pipermail/users/attachments/20130506/4ae596e6/attachment-0001.html>