On September 16, 2003 12:30 pm, Toad wrote:

> What would you prefer we do? The only way we can stop more data
> transfers being started is by rejecting queries, which is exactly what
> we are doing above! However, your bandwidth is not maxed out - so it
> would be better if the data transfers did not occupy threads. Which is
> being worked on.

What I would do is track query reject probability (QRp) for each node and 
use it in the NG routing time estimate, using something like:

Estimate = Estimate / (1-Min(QRp,0.98))

This would tend to reduce the load on nodes rejecting lots of queries avoiding
the problem when too much bandwidth sending query rejects. 

I do not think we need to save the QRp.  They are apt to change fast and we will 
quickly
determine the  busy nodes.  We _may_ have to insert code to decay QRps towards 
zero slowly so the our node eventually retries nodes that were busy.

Just an Idea,
Ed

_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to