There seem to be large scale problems at the AMS-IX. BGP sessions with
peers keep oscillating. Since their own addresses keep jumping all over
the place, it is not possible to reach anyone over the AMS-IX tech list.
I have disabled all AMS-IX peerings for the networks I manage, and I
suggest everyone who is present there looks in to doing the same.
Iljitsch van Beijnum
Iljitsch van Beijnum wrote:
There seem to be large scale problems at the AMS-IX. BGP sessions with
peers keep oscillating. Since their own addresses keep jumping all over
the place, it is not possible to reach anyone over the AMS-IX tech list.I have disabled all AMS-IX peerings for the networks I manage, and I
suggest everyone who is present there looks in to doing the same.
According to Henk Steenman (CTO AMS-IX) everything is back to normal operations. AMS-IX is still investigating what happened.
-- Arnold
I wouldn't call having to disable one link in the ring between the four
locations because spanning tree wouldn't converge otherwise "back to
normal", but yes, the AMS-IX is back to working order, as the large
number of backlogged "we are experiencing an outage" messages on the
AMS-IX mailinglist that just came in indicate...
It's very interesting to see the traffic stats at
http://www.ams-ix.net/hugegraph.html Usually, incoming and outgoing
traffic is the same. But during this problem, much more traffic went out
than came in.
Iljitsch van Beijnum
It's very interesting to see the traffic stats at
AMS-IX Amsterdam Usually, incoming and outgoing
traffic is the same. But during this problem, much more traffic went out
than came in.
Curious, as the exchange sources no traffic that shouldnt really be possible 
(unless the graphs are inaccurate)
Steve
If there is something nasty happening with STP, or traffic is flooding,
you will count the same traffic several times - once for each egress port
that it hits - hence aggregate traffic out > traffic in.
Broadcast traffic is also counted multiple times in aggregated stats.
Mike
Stephen J. Wilcox:
> It's very interesting to see the traffic stats at
> AMS-IX Amsterdam Usually, incoming and outgoing
> traffic is the same. But during this problem, much more traffic went out
> than came in.Curious, as the exchange sources no traffic that shouldnt really be possible
(unless the graphs are inaccurate)
I guess you get the same graph when you are missing data of some ports. I.e. you only have Sum(output) = Sum(input) when taking all ports into account.
Multicast and broadcast are another story as this are different counters.
-- Arnold
Unicast traffic comes in, mac forwarding table is checked and traffic
forwarded to the appropriate port. 1 byte in, 1 byte out. Port goes down
(or spanning tree sends it to blocking), all mac entries for that port are
removed, and traffic for those macs is now sent to all ports. 1 byte in,
lots of bytes out.
James