[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [RAM] Comment on draft-farinacci-lisp-00.txt (LISP)

Yakov,

>>If I can paraphrase you, Ross, doing packet forwarding at 10Mpps  
>>doesn't scare me.
> 
> 
> How many EID-to-RLOC mappings ICMP messages per second an ITR/ETR be
> required to originate/receive ?

We did a related study several years ago based on Netflow traces
collected in 1999. We used Netflow traces from two different ISPs, one
was a dialup ISP (about 6 Mbps of peak traffic) and the other a research
ISP (about 80 Mbps of peak traffic). At that time, the BGP tables in the
DFZ contained 70000 prefixes and we used the trace to look at the number
of tunnels that would have to be established towards the prefixes that
are actually receiving packets from the studied ISPs.

For the dialup ISP, there were about 2000 tunnels active (assuming that
a tunnel was active during every minute in which a packet was sent
towards the prefix) on average. For the research ISP, there were about
12000 active tunnels. The average churn for the tunnels was of about
10%. If we had tested a solution like LISP at that time, the dialup ISP
would have received on average 200 ICMP messages per minute (with peaks
of 900). For the research ISP, the average was 1100 ICMP messages per
minute and the peak about 1500 ICMP per minute.

Additional information about the traffic traces and the results may be
found in the (old) paper below

[UB00] 	S. Uhlig and O. Bonaventure. On the cost of using MPLS for
interdomain traffic. In Quality of Future Internet Services (QoFIS
2000), September 2000. Available from
http://totem.info.ucl.ac.be/publications.html

We expect that the Internet traffic has changed a lot since 1999 with
the introduction of peer-to-peer and worms. These two factors probably
increase the number tunnels that need to be established and maintained.


Olivier

-- 
CSE Dept. UCL, Belgium - http://www.info.ucl.ac.be/~obo

_______________________________________________
RAM mailing list
RAM at iab.org
https://www1.ietf.org/mailman/listinfo/ram