>Yes. Simulations of load management would be very welcome. You should
investigate the existing simulators (particularly the event-based ones),
which are in the SVN repository, I am hopeful that they can be adapted
without massive effort. You will need to more or less reproduce the current
load management system (AIMDs on the originator based on timeouts and
rejections, plus backoff to deal with individual slow peers, documented on
the wiki), and then simulate various proposals from the mailing lists for
new
systems.

Reviewing the SN repository I found
http://freenet.googlecode.com/svn/trunk/apps/load-balancing-sims/ from
phase1<http://freenet.googlecode.com/svn/trunk/apps/load-balancing-sims/phase1/>to
phase7. Are you talking about this simulations?


   - Simulations from phase1 to phase4 are only working with 3 nodes. In 6
   and 7 at least they are working with 100 nodes.
   - The simulation 4 introduces the AIMD algorithm (Peer.java).
   - In any simulation I have observed backoff process but we can consider
   that backoff is useful or it is necessary to assure by means of simulations
   that it is useful.


The proposals for extend the simulations are as follows:

   - To extend the simulations phases in large scale.
   - To compare simulations using and not using backoff.
   - To apply peer-peer token buckets to track fairness between accepting
   requests from different peers,
_______________________________________________
Devl mailing list
[email protected]
http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to