On Tue, Mar 21, 2006 at 09:10:23AM +0100, Oskar Sandberg wrote: > Daniel Stutzbach wrote: > > Kad and OverNet seem to be doing okay. They have around 1 million and > > half a million nodes, respectively. I imagine the BitTorrent DHT is > > pretty big by now, too, but I haven't played with it yet. > > Is there any more detailed information about these networks > somewhere?
They're all based on Kademlia, for which the original paper is: Petar Maymounkov and David Mazieres, "Kademlia: A Peer-to-peer Information System Based on the XOR Metric", IPTPS, 2002, http://www.scs.cs.nyu.edu/~dm/papers/maymounkov:kademlia.ps.gz I did a measurement study based on Kad, which you can find here: Daniel Stutzbach and Reza Rejaie, "Improving Lookup Performance over a Widely-Deployed DHT", to appear at INFOCOM, 2006, http://www.barsoom.org/~agthorr/papers/infocom-2006-kad.pdf Other than that, I'm not aware of much in the way of documentation of the particular implementations. Just source code. > I have played with the former a few times, and found that while it > seems to work for its purpose, it does seem to give pretty strange > and not always consistant results. Yeah, it's a little buggy. As I understand it, the guy who developed the Kad code for eMule left the eMule team before he finished debugging so it's functional but not a sleek machine. I think in the past few months someone there has starting digging into it. > I was under the impression that it used only a subset of stable > users in the actual grid, though, so a million and a half users > seems large. Yes, only non-NATed peers are in the actual grid. Meaning there's a million users in the Kad DHT, plus a whole bunch of additional users who do lookups over Kad but don't route. > (Haven't most users had been scared off these networks?) Rumors of the demise of P2P file-sharing have been greatly exaggerated. :) > What surprises me most with these networks, actually, is that they > haven't been taken down by DoS attacks. Saying this isn't an attack on > any particular system, since every system I have seen is vulnerable to > the point where it ought to take only one sufficiently motivated script > kiddie (/ national organization) with a few hundred purposely broken nodes. I spent a lot of time staring at the Kad source-code and poking and prodding it with various home-brewed measurement tools. The conclusion I came to was this: The Kad implementation had many bugs and the bugs were hard to find, so they made the system so robust that it continues to operate even though it is full of bugs. However, it is not very efficient: every node has around 700 neighbors. So I don't think a few hundred purposely broken nodes would change much. :) You should read the original Kademlia paper. I'd be interested to know if you'd think it'd be as vulnerable as you describe. > > That wouldn't be problem in Kademlia-based DHTs (nor in other DHTs > > that use parallel queries). If there's a problem, it routes around > > it. > > No, you can have similar inconsistancies that would cause hypercube > based systems like Kademlia to degenerate. Parallel queries can only > offset such problems a little, unless the network updates itself based > on which queries work well. Kademlia isn't hypercube-based, that's CAN. Kademlia does prefix-matching like Pastry, though that's not obvious from the Kademlia paper. The "XOR metric" is just a clever way of saying "find me the highest order non-matching bit". -- Daniel Stutzbach Computer Science Ph.D Student http://www.barsoom.org/~agthorr University of Oregon _______________________________________________ p2p-hackers mailing list [email protected] http://zgp.org/mailman/listinfo/p2p-hackers _______________________________________________ Here is a web page listing P2P Conferences: http://www.neurogrid.net/twiki/bin/view/Main/PeerToPeerConferences
