>Hesham wrote: > I haven't heard anyone answering my question as to why > reverse tunnelling by the MN thru the HA is so much > worse than triangular routing, > > Francis wrote in response: > => d(bidir tunnel) = 2 * d(MN,HA) + 2 * d(HA,CN) > d(triangular) = d(MN,HA) + d(HA,CN) + d(MN,CN) > d(optimization) = 2 * d(MN,CN) > and we always have d(MN,CN) <= d(MN,HA) + d(HA,CN) > so d(optimization) <= d(triangular) <= d(bidir tunnel) > and even stronger 2 * d(triangular) = d(optimization) + d(bidir tunnel) > i.e. in my poor English the cost/performance of triangular routing > is at the middle of bidirectional tunneling and routing optimization.
Yes, but does that make any significant difference for real traffic and applications given than the traffic from CN to MN goes through the HA in the triangular case? For instance assume that the CN and MN are next to each other (delay 1 ms) and the HA is 100 ms away from them (one way delay). If you route optimize TCP will see a RTT of 2 ms. If you do triangular the rtt will be 201 ms. If you do bidirectional tunneling the rtt will be 400 ms. You can look at these numbers differently depending what you want to prove. "400/201 = 2" would make your argument. But my argument is that "201/2 = 100". Thus if you care about rtt when the CN/MN are close compared to the MN/HA delay you should route optimize. The only case I can think of where tringular makes a significant performance difference is for unidirectional and latency sensitive traffic from MN to CN. I have yet to see an application which such concerns - if latency is an issue it is because you'd like the higher level entities (transport protocols or humans e.g. on a voice/video conversation) to observe short roundtrip latency. Erik -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
