Hi! On Fri, Feb 13, 2009 at 9:55 AM, slush <[email protected]> wrote: > My guess is that some nodes simply dont give me a slot for my > data, because somebody else is transfering something big, what "kill" the > bandwidth for a moment. And because there are three nodes in the line, it is > the big probability I "hit" this problem on more nodes - and I will finally > very slow responsibility.
I have another idea (which is not necessary excluding yours and is maybe just another factor/reason). Maybe Tor network is not just bandwidth bounded but also CPU bounded. For example, on my Tor node I had to increase the queue for onions size as CPU was not fast enough to cope immediately with bursts of onions (it is a 5 MB/s node). On average it is more than fast enough but there are bursts which it was not able to handle. And such bursts are probably connected with web page browsing - an user sends a lot of requests at the same time when it opens a web page. And what happens (my hypothesis) is that those go to the queue and wait for the CPU to handle them. And once they are handled they are (because of the high bandwidth) handled very fast, send back and user gets everything in the same moment (they were close to each other in the queue). I do not know what is the average waiting time for onions in the queue on my node. Maybe all this is nothing and with increasing the queue size I just increased the average waiting time from 1 ms to 5 ms. But maybe I increased it to 5 s. Who knows? Is there any way to measure it? Mitar
