Luca et al:

In catching up on the archives, I came across the thread about ntop slowing
down.  Chasing what I *thought* was a bug, I've come to realize that the
current hash scheme degrades into a linear search as the table fills up.
What it does is to use the hash as a starter into a ring-buffer and it
places the entry into the next open slot.  With a lot of entries and a full
hash, this becomes linear.

I think it is possible *post 2.0* to fix this by changing the hash to a
b-tree or similar.

Just wanted to place this into the mailing list as a reference for later...

-----Burton

_______________________________________________
Ntop-dev mailing list
[EMAIL PROTECTED]
http://listmanager.unipi.it/mailman/listinfo/ntop-dev

Reply via email to