![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
Nicholas Weaver wrote:
On Jun 2, 2008, at 10:37 AM, Joe Touch wrote:Even better would be to get some indication of where the congestion occurred. For example, if the congestion is in the 'last mile', or the other peer's 'last mile', or somewhere in between, different responses would be appropriate. For example, if my internet connection is fine, but the peer's connection is congested, I should slow down transfers from that peer, and pull more from other peers. This is, of course, easier said than done.This is not feasible for a number of reasons, including the use of tunnels and VPNs. It's possible to know whether the end systems are loaded (applications can tell each other), but path limits are hard to pin down. It's more useful to just try a few peers and see what works than to assume this kind of information is (or ever will be) available.Bulk data P2P can actually tell this if the TCP access API is suitable, by looking for correlations. If its just one flow that is seeing congestion, assume it is the other side. If multiple flows are seeing congestion, then the problem probably is at your end.
It depends on where the multiple flows go; they could as easily all experience congestion over a tunnel/VPN, even if they diverge and go to different destinations. A single connection that experiences congestion (out of multiples) does so where the connections diverge, but again that could be anywhere.
It would be useful not to over-tune P2P based on these sort of assumptions, as a result. The best you can do is split things as follows:
Congestion in the uphill path (you to the core): - all connections more likley to experience similar congestion Congestion on the downhill path (core to the servers): - connections more likely to diverge Other inferences are, AFAICT, assuming too much. Joe
Attachment:
signature.asc
Description: OpenPGP digital signature
_______________________________________________ p2pi mailing list p2pi at ietf.org https://www.ietf.org/mailman/listinfo/p2pi