Skip to Content.
Sympa Menu

bluesky - Re: Grapevine - scalability

bluesky AT lists.ibiblio.org

Subject: Global-Scale Distributed Storage Systems

List archive

Chronological Thread  
  • From: Anthony Jones <ajones AT clear.net.nz>
  • To: "Global-Scale Distributed Storage Systems" <bluesky AT franklin.oit.unc.edu>
  • Subject: Re: Grapevine - scalability
  • Date: Wed, 12 Dec 2001 07:23:06 +0800


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

> > I'm not saying it cannot work, but personally I will keep trying to get
> > the correct emergent properties out of networks consisting of nodes
> > behaving under a simple set of rules to the grave sooner than give in to
> > the nightmare of trying to keep this sort of network consistant in the
> > wild.
>
> I don't believe that CAN/Chord will ever make sense for the kind of
> transient relationships you get in public nets like gnutella, but I think
> that CAN/Chord make sense when the social environment is right. In a
> business environment social pressures to do the right thing should be
> enough to get churn down to a rate that's slow enough for the
> reconstruction logic to work.

Every P2P strategy has its strengths and weaknesses. The Grapevine project
has similarities to CAN but it also has differences. What we (the open P2P
community) have to do is choose a strategy that we think will work and try
it. It will either work or not. Either way others can learn from the
experience. We are all both competing against eachother and all working for
the same cause. Either way information will be free to flow on the Internet
again one day.

I am confident that Grapevine will work and we are all (Grapevine developers)
going to give it our best shot. The reason I think it will work is that it
sits between the self organizing of Freenet (which I believe doesn't work
well enough) and the structured nature of CAN (which I think is too
structured and complex).

It is just as important for Freenet as it is for Grapevine for nodes to stay
up. The more time they stay up the better the network will work as a whole.
Key things you need to optimize to make it good for the user are:

1. The overhead (background traffic) needs to be reasonable.
2. The efficiency needs to be high - i.e. the network shouldn't have to
transfer 5K for every 1K downloaded.

If you don't do these then the node traffic will be high and the user will
shut the service down.

Anthony

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE8FpVdhwVaoilFPn0RAgxyAJ4woN2M2FUR3KygieKdVXuLHIuqaQCfaB8j
Xm5sBtAItXGl9USi5ZLUg7w=
=hoyP
-----END PGP SIGNATURE-----




Archive powered by MHonArc 2.6.24.

Top of Page