Will P2P Scale?
from the ask-Techdirt? dept
Rocker writes “With the peer-to-peer rage and as some predict “the next big thing”, I was wondering the obvious: How will the millions of future P2P users find each other without a central (and vulnerable) host such as Napster? It seems obvious that even emerging technologies like Gnutella need an option other than sharing IP addresses. I’m surprised that more of these “up and coming” P2P ventures don’t use domain names suited for dynamic IP addresses like the service provided by DialupDNS and others. I’m curious what the readers of Techdirt think about this topic and what solutions they see.” Slashdot has been pretty successful with their “Ask Slashdot” feature, so maybe this is a test of the “Ask Techdirt” feature. Of course, our readership is a bit quieter than theirs… Though, if you have any opinions on the question, start typing.
Comments on “Will P2P Scale?”
maybe I'm wrong...
But I thought instead of a “central server” they’d have mirrored-central-server-like servers. But yes, P2P is inherently not as scalable. But I also thought the best application of P2P was for “communities” and not “universal” file storage… Anyway, I think some hybrid of napster/gnutella will be hobbled together at some point to do the job adequately. There are no perfect solutions to really hard problems, but we can approximate answers as much as we want….
Re: maybe I'm wrong...
There are two reasons why not to have a central server makes sense in p2p architectures. The first one is (as with Gnutella or Freenet) not to be attackable, to have no point of failure. In other words: Allow “illegal” p2p activites which cannot be controlled or shut down. That is in my eyes the most important reason. Then there is another one, stating technical advantages linked to not having a central server, saying that a central server less p2p architecture is more efficient. However, even early p2p developers realize by now that this is less and less the case the bigger the network is. As scalability issues have become the major obstacle of p2p, the general trend reversed towards more centralism, and most developers by now agree that the most efficient technological p2p architecture from a current perspective utilizes a high number of central services. Sticking to fully distributed p2p would then only make sense for the first reason. Or does anybody have another opinion on this?