Re: Re: Re: DHT vs Gnutella

Am Mittwoch 06 August 2008 22:09:32 schrieb Sam Berlin:
> > Please stop the quarrelling.
>
> You coulda quoted Raphael there, atleast.  I'm just asking the questions! 
> :-)

Sorry, I just answered the last post in the row :) 

Heats sadly builds up gradually... 

I well understand that people can get sarcastic at times and that it hurst to 
be hit by sarcasm, but at the moment I'm in a position of not being hit, so I 
can at least try to take a bit of the heat out of the discussion. 

I hope I didn't botch that exceptionally :) 

> > How much data do you really need to keep?
>
> You need to keep each peer's QRP table -- both leaves & ultrapeers.
> Depending on the implementation, the data can be very memory
> intensive.  QRP tables are the single-largest memory hog in LimeWire
> right now -- before we can even think about increasing the number of
> tables, we need to reduce the memory overhead.

Would you like to have a look at the phex code? 
I think I can at least locate the corresponding sources: 
- 
https://phex.svn.sourceforge.net/svnroot/phex/phex/trunk/src/main/java/phex/common/QueryRoutingTable.java

I hope that one helps you. 

> I would like to think that the target audience of LimeWire is a little
> less techno-savvy than that of Phex or gtk-gnutella (no offense
> intended at all).  

No offense taken :) 
Phex is written for more tech savvy people, so it's only natural to have a 
different group of users. 

> Given that, the program really should stay limited 
> in its memory requirements.  First and foremost it's for the user --
> if the program starts eating away at the computer's resources, the
> user isn't terribly happy.

Did you increase the amount of memory you allow LimeWire during the last few 
years? 

Ram in new computers became far cheaper, and every new computer now has a few 
Gigabytes of Ram available, so taking a hundred MiBs when you're an UP 
shouldn't hurt so much anymore. 

I think the Ram available at least doubled, and in Phex we realized that the 
program behaved far better (under GNU/Linux and OSX at least) if it got a bit 
more memory :) 


There's one question I want to ask anew: 

> Do you have statistics, how full a table normally can be until searches 
travel to too many hosts (for example until searches are forwarded to more 
than 1/3rd of the connections - which is quite good if it's the step farthest 
away from the leafs, I think)?


Best wishes, 
Arne

-- My stuff: http://draketo.de - stories, songs, poems, programs and stuff :)
-- Infinite Hands: http://infinite-hands.draketo.de - singing a part of the 
history of free software. 
-- Ein Würfel System: http://1w6.org - einfach saubere (Rollenspiel-) Regeln. 

-- PGP/GnuPG: http://draketo.de/inhalt/ich/pubkey.txt


[Non-text portions of this message have been removed]