> I think the question is, can we do what we want to do within > the bandwidth that our target users have available to them... > such as say a 56k channel, 128k, 512k, etc. If we're using > 32 of 56k for maintenance, that may leave little left for the > signal you want to send through.
We went through that discussion several years ago during the design of PNRP. One of our goals was to ensure that if two nodes were on the same "area" they could find each other even if that area was cut-off from the rest of the Internet. It turns out to be an interesting challenge, which leads to some form of minimization of the maintenance traffic. There are fewer and fewer candidates for "routing peers" as you go through the levels of your routing table. At the first level, for each slot, if you have N nodes in the network and K slots per level in your routing table, there are N/K candidates per slot. Then, N/K^2 at the next level, etc. Chances are that you can populate the first levels with nodes that are in your area. This has the effect of minimizing the maintenance traffic, and making it easier to contain it. But at the most granular levels, your adjacencies will be all over the Internet. There is definitely an effect on transcontinental links. Suppose each node sends a mere 100bps average of such traffic. If you have 100 million nodes, that's 10 Gbps through the intercontinental link, just for maintenance. Now, you may say that's a nice problem to have, because it means wild success. But it still is a problem. -- Christian Huitema _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
