|
Peer to peer scaling....
|
Kevin Jamieson
|
Feb 14, 2001 17:33 PST
|
Ok, this link has been posted to both Slashdot and Memepool, so far all I
know it could be old hat by now:
http://www.darkridge.com/~jpr5/doc/gnutella.html
It's a paper regarding the technical issues involved in making Gnutella, our
Favorite File Sharing Service That Is Named After A Food, scale effectively
to potentially millions of users. Well, actually, it states that it won't
scale at all much past the thousands. It involves quite a lot of math and
some very well thought out guesstimating based upon the numbers involved.
Now, there might be some bias in the article itself - the author of it did
help to build Napster, after all.
Overall it provides an interesting and fairly technical look at the
obstacles that have to be overcome regarding massive peer-to-peer
filesharing services.
Here's my take on it: Gnutella to me would still be useful, even in its
current form. It would be nice to be able to set up a multi-platform
filesharing pool between myself and other close friends/associates. Like
BBS's from the old days or 'warez groups', Gnutella can be used among select
groups of people to trade exactly what they want. Take, in this case, mp3's
- I know people who easily have ten gigs of mp3's, each. Now, there
probably would be some overlap, but if we all hooked up onto a small private
gnutellanet there could be probably 100gb or so of unique mp3's at our
disposal.
The main problem with scaling discussed in this article is the overall
bandwidth required to maintain the overhead of the search transactions... if
you kept your gnutellanet to say, less than 50 people, and had some baseline
requirements (i.e., you have to have DSL, no dialup - you have to have a
machine that can be up for days at a time so as to get rid of the
'transience' factor) it could be quite a stable and overall efficient
environment to work with.
The value to me personally in a large network like Napster is it increases
my odds of finding obscure tracks by obscure bands - the idea that somebody
out there, at least one person, must be interested in what I am. Maybe the
people on Napster who are all interested in the same artists (Say, popular
San Francisco club DJ's) will band together to form their own small
gnutellanets to trade tracks and keep in touch. By making small,
specialized networks that have limitations on scaling, it could work out in
present form.
Of course, that introduces the concept of some authority, centralized or
not.... Which is another kettle of fish, altogether. :)
=-=-=-=-=-=-
I can't remember if this was posted to the list before, but here it is for
reference anyways - Salon did a piece a few days ago about online marketing
via Napster/peer to peer file exchange services..
http://www.salon.com/tech/feature/2001/02/09/napster_parasites/index.html
Kevin
|
|
 |
|
