Stupid server semantic argument (was: Non Linux but network tech question)

[Bill McGonigle]

On Jun 24, 2007, at 22:05, Ben Scott wrote:

> This also destroys the
> anti-leech protocol arguments; implementing such just decreases the
> popularity of the protocols.

I've only done protocol work with Bittorrent, but as I understand it,  
it's quite popular.

Bittorrent uses economic and game theory models to determine who gets  
the bandwidth.  In the official protocol implementation, Bittorrent  
uses a modified version of the Tit-for-Tat solution to the Prisoner's  
Dilemma.  Every 10 seconds it reevaluates the sharing ratio of its  
[default 4] peers, using a 20-second moving average of DL/UL ratio,  
and will 'choke' peers which perform badly.  Additionally, every 30  
seconds, it uses an extra client to opportunistically unchoke a new  
peer [default random selection] to measure its adeptness at  
cooperation (similar to the opening move in the Prisoner's Dilemma).   
If the client turns out to reciprocate, and is superior to a client  
in the existing pool, it will replace the poorly performing client.   
This system comes close to achieving a pareto efficiency for the  
members of the swarm.

There are alternate methods for opportunistic unchoking other than  
random selection at a 30 second interval on a single thread.  I've  
done a bit of work with network closeness as a means to minimize  
transit costs of Bittorrent.  Some guys at Stanford have done a pure- 
efficiency play which maximizes the benefit to a given client, and  
might even be a decent way to achieve low cost on the entire swarm.   
This algorithm comes closest to looking like a leech, but the benefit  
appears to be largely reciprocal.

That's just for the swarming, or P2P, phase of a transfer.  Once it  
has the complete file it can begin seeding, at which point is has no  
particular need to disfavor leeches.

-Bill

-----
Bill McGonigle, Owner           Work: 603.448.4440
BFC Computing, LLC              Home: 603.448.1668
bill at bfccomputing.com           Cell: 603.252.2606
http://www.bfccomputing.com/    Page: 603.442.1833
Blog: http://blog.bfccomputing.com/
VCard: http://bfccomputing.com/vcard/bill.vcf