On Wed, 18 Aug 2004, Matthew Williams wrote:
If we have the address space available, shouldn't we just go ahead and help him out? Providing he makes the appropriate references when the results are published, of course ;)
As we discussed this morning: yes, let's just do this for a month or two. Henk
Cheers, Matthew
-----Original Message----- From: ris-int-admin@ripe.net [mailto:ris-int-admin@ripe.net] On Behalf Of RIS Sent: 16 August 2004 12:40 To: ris-int@ripe.net Subject: [ris-int] Fw: NCC#2004081420 [RIPE beacon request!]
Hi guys,
What do you think about the following request? It looks he is asking to extend our anycast failover scenario to more RRCs.
After discussing with him what kind of patterns he needs, we can use same prefix or we can use one of the free Beacon Prefixes.
Arife
----- Begin forwarded message -----
Date: Fri, 13 Aug 2004 20:39:14 -0400 From: "Hitesh Ballani" <hitesh@cs.cornell.edu> To: <ris@ripe.net> Subject: NCC#2004081420 RIPE beacon request! Message-Id: <81224EA6DDA07C43BC7555E32401029E054219@EXCHVS2.cs.cornell.edu>
Hello,
I was wondering if you could spare a beacon for the following experiment: At the moment, one of your beacons (195.80.233.0/24) at advertised from 2 locations - LINX and AMS-IX. An interesting question is to study anycast fail-over time from an end user perspective - i.e given that the advertisement is withdrawn from AMS-IX at a particular instant, how long is it before convergence is reached and packets from a node, which were earlier being routed to AMS-IX, are routed to LINX. I tried running this experiment from a bunch of nodes on Planetlab but I have not been able to locate any node whose packets are routed to AMS-IX and so, I am never able to analyze the convergence time (for that matter, even nodes in Netherlands have their packets routed to LINX). Here, "packets being routed to LINX" implies that periodic traceroutes from my Planetlab sites show so. This is important to me as I am currently working an a proxy based deployment model for IP anycast and so anycast failover time through BGP is critical for me (see http://www.cs.cornell.edu/People/francis/pias.pdf and http://www.cs.cornell.edu/People/francis/pias/pias-worlds.pdf for details!). As a matter of fact we are in the process of obtaining 3 /22 blocks from ARIN for this very purpose - however, given the long drawn nature of that process, I thought it would be nice if I could use your advertisement infrastructure to answer some niggling questions which are critical for the success of our proposal.
I was wondering if it would be possible for me to get a beacon to be advertised from more locations (ideally, from America(RRC08), Europe(a number of locations), and Asia(RRC06)). While a beacon being advertised from 2 locations would suffice for the above experiment, but advertisements from more than 2 locations would allow me to look at other stuff such as - How good is IP anycast at selecting nearby locations? Ideally, when a node sends out a packet to a anycast location, it should be routed to the destination which is closest - but with the policy based nature of BGP, it is not clear if it does. Such an experiment would make more sense with greater number of advertising locations. I tried determining answers to these questions using the anycasted root servers and the servers belonging to the AS-112 project but the hierarchical nature of their deployment prevents me from coming to any conclusions. While the above description may sound as if I am looking for just an end-user analysis, but I will also be looking at BGP updates for the anycast prefix to determine if advertising a prefix from many locations has an adverse impact on the visibility of the prefix due to reasons such as route flap dampening.
If the above description sounds terribly confusing, I will be glad to provide additional details about the experiments I plan to run. Also, if such an arrangement is possible, I will plan out the detailed A/W pattern and the sites involved.
Thanks, Hitesh ------ End forwarded message ------
------------------------------------------------------------------------------ Henk Uijterwaal Email: henk.uijterwaal(at)ripe.net RIPE Network Coordination Centre http://www.amsterdamned.org/~henk P.O.Box 10096 Singel 258 Phone: +31.20.5354414 1001 EB Amsterdam 1016 AB Amsterdam Fax: +31.20.5354445 The Netherlands The Netherlands Mobile: +31.6.55861746 ------------------------------------------------------------------------------ Process and Procedure are the last hiding place of people without the wit and wisdom to do their job properly. (David Brent).