With the Napster debate still raging and P2P networking still flavour of the month, IT managers might be alarmed to realise that they could already have examples of these radical P2P applications thriving within their own four walls. In obscure departments and branch offices, far from the watchful eye of the network management department, your staff may even now be freely exchanging files without the control of a central server. I'm afraid it's true - good old Windows 95 and Windows for Workgroups were offering P2P networking before Napster was even in nappies.
Though P2P has hit the headlines lately, the concept of a network without a central server isn't really new at all - the fact is that most local networking used to be that way. So the issue for IT professionals is, if it was once seen as a good idea to leave that P2P model behind in favour of central network control, why should we be interested in P2P now?
The Internet, as Oracle puts it, changes everything. It's certainly changed things for P2P. Windows for Workgroups may be able to link everyone in the accounts department, but P2P projects running over the Internet can link participants across several continents.
Around 50 million people use the Napster network to swap music over the Internet. Over 2.8 million participants have signed up to help with Berkeley University's [email protected] project, which harnesses the power of home computers to
But what these projects have in common is not so much a shared technology as a shared community spirit. Swapping music is cool; so is hunting for aliens over the Web, or finding a cure for a killer disease. It's a buzz to watch your PC processing yet another Fast Fourier Transform and reflect that maybe this particular SETI work unit will be the one that will prove that there is intelligent life out there in space. All this is undoubtedly an exciting new way of making the Internet work for social goals. However one hates to say it, but is there any money in it? Is P2P a viable commercial, as opposed to social phenomenon?
Lots of people seem to think so. Sun Microsystems, Intel, Exodus Communications and a group of ex-Netscape staff are among those involved in new ventures intended to exploit the power of P2P, and in this programme you'll find out about others. There are existing P2P companies too; Napster, of course, but also companies such as Entropia, set up four years ago to carry out network computing services using recycled PC power.
Existing P2P ventures work because participants have a vested interest in getting involved. In Napster's case, people want free music. In Entropia's case, they want the feelgood factor of supporting good causes such as fighting AIDS; of "making Internet history" as the company grandiosely puts it on its Web site. But what happens if you're not helping make history, just making money for some Internet company? Will P2P companies offer people payment for their participation and if so, how viable will the business model be then?
The question of cost rears its ugly head
To query the economic validity of P2P may seem naïve when the global investment community has spent two years merrily pouring billions into Internet companies that never had a cat in hell's chance of making a profit. (And indeed, have put money into Napster, which has no obvious way of making a profit either.) But to get the participation that these kinds of P2P projects need, it's not gullible venture capitalists who have to be convinced - it's individual computer owners. Amid the P2P frenzy, it's sobering to note that one startup, Popular Power, folded just a year after its launch.
Not all P2P projects involve using free time on home PCs, of course. As part of this P2P programme, we'll be looking at ways in which P2P could be used in the corporate environment. But when we move into the corporate world, the question of cost rears its ugly head once again.
As we all know, it's cost of ownership that eats up IT budgets, not the cost of buying CPU power. And to be commercially viable, even P2P networks have to be managed - indeed, they arguably need more careful management than conventional centralised ones.
Gartner Group has defined five different types of P2P applications. Very few of the applications mentioned to date fall into its pure "atomistic" category, where communication isn't mediated by servers but nodes on the network communicate as equal participants. The rest require central coordination to a greater or lesser extent. Napster uses millions of home PCs for storage, but the whole system is managed by a central server that holds lists of songs and users. SETI uses home PCs for processing, but work units are doled out and results collated centrally. The SETI project is funded by donations. Who's going to pay for management of commercial P2P networks and what are the benefits?
Clay Shirky, partner at investment firm Acceleratorgroup and nascent P2P guru, has described Internet-connected PCs as "the dark matter of the Internet" and argued that P2P's strength is in channelling that dark matter for productive ends. Fair enough when you're alien-hunting in the not-for-profit academic world, but it's not as if PC power is exactly a scarce commodity in the corporate environment. Indeed, Microsoft has often defended its CPU-gobbling bloatware on the basis that, hey, PC price-performance is tumbling and there's no need to write lean, mean applications any more.
That being the case, isn't it more than a touch ironic that Microsoft's old ally Intel is now extolling the virtues of P2P as a way of making the most of your distributed computing power? Our programme of P2P content over the coming days will provide the answers to these questions.
Five things you might not know about peer-to-peer computing
- Gartner Group reckons that by 2002, more than half of global Internet users will regularly sign on to at least two P2P Internet applications.
2. There is about 10k terabytes of storage on the Internet which P2P applications could tap into.
3. P2P company Entropia claims its distributed network can generate processing rates of 1 Teraflop per second, 24 hours a day, seven days a week.
4. A recent Jupiter/Media Metrix survey of Web site hits in January and February said the two sites with the highest traffic in the first two months of 2001 were the Napster site - and the US Inland Revenue Service homepage.
5. Though Napster isn't strictly peer to peer in structure, truly distributed P2P models do exist; the Gnutella file distribution system has a distributed index and can work without a central server.
Useful P2P links
Share and Share alike
Supercomputing on a shoestring
CW360 P2P Glossary