Ruud Javi wrote: > Some questions: > - Is it misrouting to route A-C-D-E, instead of A-B-E in above example. > You would need one additional node, but you prevent the use of a slow > bottleneck node (that in the current situation would probably back-off > from time to time). > > - This mechanism would only take time into affect between nodes. Does it > also take time on a node itself? It could be added easily into the formula. > > - Does this mechanism have disadvantages compared with the current routing? > > - Is it hard to determine travel times between nodes? How long does a > transfer take on average and how long does a request take on average?
It's easy enough to determine the travel time to your immediate neighbours, but if misrouting adds an extra hop to the path (A-C-D-E instead of A-B-E), the overall latency might increase even if the latency of each hop is lower. The overall bandwidth usage would also increase. Cheers, Michael
