P2P-Zone

P2P-Zone (http://www.p2p-zone.com/underground/index.php)
-   Peer to Peer (http://www.p2p-zone.com/underground/forumdisplay.php?f=5)
-   -   P2P Inefficient But Improving, Researchers Claim (http://www.p2p-zone.com/underground/showthread.php?t=13587)

JackSpratts 12-10-02 09:34 AM

P2P Inefficient But Improving, Researchers Claim
 
Mike Martin, sci.NewsFactor.com

Internet peer-to-peer file-swapping programs that strike fear into the hearts of recording artists and distributors -- such as Gnutella and Napster -- may not be such terrifyingly efficient instruments of intellectual-property appropriation after all.

An exhaustive study of Gnutella has convinced University of Chicago researchers that Gnutella's architecture does not match the underlying topology of the Internet. They believe Gnutella travels the Web on inefficient wheels -- like a car without snow tires in a blizzard, or a highway-touring bike on a rough mountain trail.

Nutty Wiring in Gnutella?

"The way the Internet is wired makes a message going from New York to San Francisco pass through Chicago, for instance," researcher Matei Ripeanu told NewsFactor. "The way Gnutella is 'wired' -- without regard to the underlying Internet -- might make a message from New York to San Francisco pass through Tokyo."

Gnutella nodes, called 'servents' (sic) by developers, perform tasks normally associated with both servers and clients, Ripeanu explained. These nodes provide client-side interfaces through which users can issue queries, view search results and accept queries from other nodes.

"The nodes are also responsible for managing the background traffic that spreads the information used to maintain network integrity," he said.
http://sci.newsfactor.com/perl/story/19647.html

the article also hints that network destabilization may be relatively easy...

- js.

Mazer 12-10-02 02:48 PM

Someday someone will make a p2p app that follows geographic lines but I don't think it's completly necessary. What if the route through Tokyo is faster? Eventually routers will be so fast that even small ISP's will have the bandwidth that the internet backbone has today. The fact is that there are internet outages that might cut off one or two routes through a geographic area and a good p2p app would work around the fault. If it strictly followed geographic lines then an outage would fragment the p2p network.

zombywoof 12-10-02 02:57 PM

It still beats shelling out bucks in the record store though.

pod 13-10-02 01:01 AM

Re: P2P Inefficient But Improving, Researchers Claim
 
Quote:

Originally posted by JackSpratts
Internet peer-to-peer file-swapping programs that strike fear into the hearts of recording artists and distributors -- such as Gnutella and Napster
Maybe the guy should have taken some English classes along with the journalism ones... :p I had no idea Gnutella and Napster are recording artists and distributors. :a:

Quote:

"The way Gnutella is 'wired' -- without regard to the underlying Internet -- might make a message from New York to San Francisco pass through Tokyo."
So might a regular Internet message, technically. Even currently there are all kinds of business and political reasons that cause packets destined to my neighbour (who's on a different ISP) go half-way around the world instead of some local interchange point.

Makes me wonder if those 'researchers' got a grant for this 'study'. :sus:

jaan 13-10-02 04:15 AM

Re: Re: P2P Inefficient But Improving, Researchers Claim
 
Quote:

Originally posted by pod
So might a regular Internet message, technically. Even currently there are all kinds of business and political reasons that cause packets destined to my neighbour (who's on a different ISP) go half-way around the world instead of some local interchange point.
correct. however, the article has a point: gnutella ignores the geographical locations of its nodes, making it on average less efficient than the underlying IP network.

the article/research gets it wrong though by implying that all current P2P architectures are similar to gnutella, and applying their results to other P2P networks.

plus, saying that open source protocols are inherently more flexible than closed source ones, is not true at all. as we've seen with gnutella, new protocol features must be agreed upon and implemented by the numerous independent client developers before they are to succeed.

- jaan

pod 13-10-02 09:25 PM

I know, I know. The research may be correct, but it's still stating the obvious. Anyone who's been on Gnutella recently knows that it's definitely not the future.

Speaking of the future... There is some really cool research happening WRT P2P network architecture. The goal is, of course, to have the nodes very well connected, but keep the chatter down. I think a good start would be to determine for every new client connecting who he's closest to, network-wise, and have him join the network over there. And I do believe someone is trying to make a geographically aware client, but my opinion of that is to just stick to optimizing network topography. LIke I said, the guy could be next door, but that's not saying much.


All times are GMT -6. The time now is 03:45 PM.

Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
© www.p2p-zone.com - Napsterites - 2000 - 2024 (Contact grm1@iinet.net.au for all admin enquiries)