In this example, we will make a computation between using and not using the extension to determine its effectiveness.
The test consists of downloading all the metadata to make an extensive listing, but first we must make a cache cleanup.
# yum clean all
...
Cleaning repos: _local fedora fedora-chromium livna rpmfusion-free
: rpmfusion-free-updates rpmfusion-nonfree
: rpmfusion-nonfree-updates updates
Cleaning up Everything
507 delta-package files removed, by presto
fastestmirror must be enabled in the configuration file and then starts the measuring of time required to make a comprehensive list:
# time yum list available
...
zzuf.x86_64 0.13-1.fc13 fedora
real 3m19.740s
user 0m16.017s
sys 0m1.269s
Now the experiment will be repeated disabling fastestmirror, but first, a cache cleanup is made.
# yum clean all
# time yum list --disableplugin=fastestmirror available
...
zziplib-utils.x86_64 0.13.49-8.fc12 fedora
zzuf.x86_64 0.13-1.fc13 fedora
real 3m24.678s
user 0m16.125s
sys 0m1.270s
The gain has been small. The details of how it measures the speed of the repositories is not known, but the truth is that it does not seem to be very effective. Even though, according to the experiments, it works according to its own logic, ie measuring all the servers and using the fastest according to that measure, hence there is potential for doing it better . One of the tests was to check whether this extension acts after assembling the list of servers regardless of the failovermethod method selected in the repositories configuration, ie, it works regardless of the order of that list, re-ordering it according to the speed measured and choosing the "fastest".
