On Wed, Dec 03, 2014 at 12:15:16PM +1300, Simon Lyall said:
Personally I always had a sneaking suspicion that Citylink never
released graphs because the actual amount of traffic was tiny.
Probably not true but how could I tell?
For a long time, particularly on WIX, there was a mechanical problem of
defining what an exchange port, and identifying IX traffic. People
pushed traffic into a single shared vlan, some was transit, some was
peering, some was point to point private stuff. A graph of all the
ingress data in that vlan wasn't particularly meaningful as a way of
describing the peering activity, since it had (literally) hundreds of
ports doing things unrelated to BGP.
Once the process of draining the swamp was underway - splitting the
single vlan into transit vlans per ISP and private vlans for point to
point (so that the IX was all that was left in the original vlan), there
was a period (of a couple of years) where an aggregate graph of the
original vlan would have suggested the amount of exchange traffic was
falling. During that period, there definitely was a lack of enthusiasm
for publishing aggregate ingress data, since the graph trend didn't
match reality (at least, reality as we assumed it to be - we didn't have
much other hard data on IX trends).
Cheers
Simon
