http://www.newscientist.com/news/news.jsp?id=ns99992861
New P2P network funded by US government 14:28 01 October 02 NewScientist.com news service A team of government-funded US scientists is building a Peer-2-Peer (P2P) network that they say will solve technical problems with existing P2P networks, such as Gnutella and Kazaa, and might even one day supercede the web. The network, dubbed the Infrastructure for Resilient Internet Systems (IRIS), will speed up searches and information transfer over the internet, and aims to foil "Denial of Service" attacks by hackers - in which a web server is swamped with requests for a page until it crashes. On the web, information is usually kept on a central server. But in IRIS information will be duplicated as its popularity increases, thereby sharing the load over many machines. Shifting the information means the information can dodge an attack. "It will stop servers from crashing under Denial of Service attacks," says Hari Balakrishnan, a computer scientist based at MIT and a principal researcher on the project. Design criteria P2P networks connect together a large number of different computers, pooling their resources. They provide an efficient way of storing and accessing large amounts of data, as demonstrated by the popularity of music file sharing networks such as Kazaa. However, P2P networks are de-centralised, i.e. there is no central server that keeps tracks of all the information in the system. This means searching for information on these networks is slow and cannot guarantee to find information. IRIS is being designed specifically to solve these problems. Its three design criteria are to guarantee: � that as long as there is no physical break in the network the target file will always be found; � that adding more information to the network will not affect its performance; � that machines can be added and removed from the network without any noticeable adverse affects. "There is no single network that meets all these three properties as yet," Balakrishnan told New Scientist. Logarithmic increase The new search algorithm that Balakrishnan and his colleagues are developing will find a file on IRIS quickly. Crucially, the search time will only increase logarithmically compared to the increases in the amount of stored data. So the size of the network can grow, and the search time will not increase dramatically. The project will be developed over the next five years by researchers from five institutions, including MIT and the University of California at Berkley, who have jointly received a $12 million grant from the National Science Foundation to develop IRIS. Balakrishnan hopes that IRIS will eventually be adopted globally as a default standard for information exchange. "We think IRIS should be used for more than just file sharing," he says. Bad things The first application will be a distributed version of the web. This raises the prospect of it being very easy to published information anonymously, for example, pirated music and video. But he does not believe this should curtail his research. "How do you prevent people from doing bad things? I don't think this is a technical problem," says Balakrishnan. In fact his team is developing algorithms precisely to thwart the censorship or control of information on IRIS. "People are working in our team to prevent removal of information," he says. "I am not interested in censoring the publishing of information." _______________________________________________ http://www.mccmedia.com/mailman/listinfo/brin-l
