P2P Networks (TCD 4BA2 Project 2002/03)

Intoduction

1. Historical Development

Authors

Overview

Past

Present

Future

Conclusion

2. Music and P2P

3. Copyright and P2P

4. Napster

5. GNUtella

6. YouServ

7. Freenet

8. P2P Search Engines

9. P2P Routing

10. P2P Security

Readers Guide



Historical Development

 

Authors:

Andrew Brosnan brosnaa@tcd.ie
Tarun Maitrat maitrat@tcd.ie
Andrew Colhoun colhouna@tcd.ie
Bob MacArdle macardar@tcd.ie

 

Overview

Before discussing the history of peer-to-peer computing (P2P), we must first decide what it is. According to Whatis.com, P2P is "A communications model in which each party has the same capabilities and either party can initiate a communication session". So we could define P2P as direct communication or collaboration (mostly file-sharing) between computers, where none are simply client or server, but all machines are equals - peers. In that case, P2P is at least 30 years old1. With this definition, communication between two servers is P2P, and Napster (strictly speaking, with it's central directory) isn't. It could even be argued that telephones and email are P2P.

However we will, with Clay Shirky's help2, use the epithet P2P to describe systems whereby ordinary nodes - in particular desktop PCs - at the edge of a network are given substantial independence, and which handle the variable addresses of these nodes. What makes P2P unique is not two nodes talking to each other as equals, but the type and (virtual) location of the nodes. Ordinary PCs, previously little more than web-page viewers, become active participants in the Internet, despite their lack of fixed IP addresses.

A more technical definition of P2P was put together by Dave Winer of UserLand Software. He suggests that for a system to be P2P it should have the following seven characteristics.

The basic idea is that two computing devices (peers) share resources and information with each other, with both acting as a kind of mini-server, that is to say neither is specifically a client or server. As there is no need for any central web-server, problems with controlling the information were soon to follow. A good quote remarking on this was made by Katherine Mieszkowski of Salon - "P2P is a particularly comical new coinage for a business model since the phrase starkly points out that there's no middleman - so how can anyone possibly make any money".

1. www.peer-to-peerwg.org/whatis/index.html
2. www.openp2p.com/pub/a/p2p/2000/11/24/shirky1-whatisp2p.html?page=2  

Past

What came before P2P?

Before we had P2P, one of the big favourites for companies to use was B2C. B2C stands for "Business-To-Consumer".

The P2P Overnight Explosion

The history of P2P cannot be recounted without reference to Napster, the software that brought P2P into the mainstream. Napster became the single most popular P2P application literally overnight.

It all started in 1999 when an 18-year-old college student, Shawn Fanning, left university after the first semester of his freshman year to work on the Napster software. Frustrated with how difficult it was to locate and swap digital music files online, and aware that many internet users had many of these files sitting on there hard disks, he designed a piece of software which combined the instant-messaging system of IRC, the file-sharing functions of Microsoft Windows and Unix, and the advanced searching capabilities of various search engines, he created Napster (named after his college nickname). It became the fastest growing software online, of all time.

Napster's winning idea was in giving P2P, for free, to the masses. At the same time as Napster was released three factors greatly increased it's mass popularity - higher bandwidth, more powerful desktop processors and cheaper storage. Add to this the fact that for many American homes, Internet access is flat rate, which led to people leaving their computers on 24-7 downloading music. All a person had to do was download Napster, search for the file he wanted, and then watch all there favourite music filter down onto their machine. However the huge growth of Napster was a great concern for record companies around the world, and on April 13th 2000, nearly after its release, Rock Band Metallica sued Napster for copyright infringement. After a lengthy legal battle Napster reached an agreement on September 25th 2001. The settlement says that Napster will pay 26 million to songwriters and publishers as well as a percentage of the money it takes in when it starts its pay service in early 2003.

Since the Napster case, many other P2P programs have surfaced including Gnutella, KaZaa and WinMP. And P2P has moved beyond file-sharing, to such projects as SETI@home at UC Berkeley, which pools the spare processor power of the participants rather than their MP3 collections.

 

Present

The growth in the number of P2P technology users has been exponential since the birth of Napster. The vacuum left by Napster's demise has been filled by numerous other companies/applications, with Kazaa and Gnutella dominating the market. It has been reported that these two account for between 40% and 60% of all traffic on the Internet (1). The massive growth has been accompanied by massive strides in the development and understanding of P2P technologies. Unlike the centralised techniques of Napster, the new applications are adopting a decentralised approach, making them harder to police. This has led to a lot of concern about the lack of central leadership and control.

In recent years we have seen P2P technologies being embraced by large companies trying to tap its vast potential. Two of the more notable examples of this are Deloitte & Touche, and Intel. Central databases are no longer required when using P2P. This means it can be less expensive and far easier to scale. Intel has been using P2P since 1992, thus avoiding the need for a large server. Traditional databases are still commonplace today, but as increasing numbers of companies follow the example of organisations like Intel, such databases could be overtaken and replaced by P2P.

A significant turn in the development of P2P occurred recently when Microsoft announced plans to invest $51m in a company called Groove Networks. At the forefront of this company is Ray Ozzie, the inventor of 'Lotus Notes'. Groove Networks is implementing a hybrid technique in its development of P2P. This means it utilises both centralised and decentralised techniques in order to get the best of both worlds.

The development of P2P technologies has been hindered by a number of legal issues. The newer companies, such as Morpheus, are seeking to learn from Napster's mistakes and get legal protection. They are arguing that they are not responsible for any illegal activity perpetrated using their software. There are clearly a number of legitimate legal uses for P2P and this should safeguard its future. The only problem is how to stop illegal file sharing (movies, music etc.). At the moment no one has the answer, but the search is ongoing.

Since they are most at risk, record companies are currently looking at ways of deterring people from using copyright-busting file-sharing networks. The open nature of these systems allows record companies to attack them from within - creating their own fake users who provide bad quality or confusingly-named data, or overloading the network with queries of their own. This technique was put forward by two students from Washington University (2), who outlined it in a white paper. However, the legality of this process is itself being questioned. The students also suggested that users be randomly sued, or that heavy users, P2Ps equivalent of dealers, be targeted. It seems, at present, this would be the best way to deter people from illegally sharing files.

(1) www.vnunet.com/News/1134977
(2) http://students.washington.edu/achen/papers/p2p-finalpaper.pdf

 

Future

It's hard to know what the future of P2P is until the legal situation becomes clear. Companies will have to expend significant resources to protect themselves against lawsuits. There will always be illegal file-sharing and copyright infringement, but these will be at the periphery. Any one system that gets too large will presumably suffer the same fate as Napster, unless it is so decentralised and anarchic that no one can be held accountable. A lot depends on the legislation drawn up by governments in response to P2P. Whether the burden of preventing copyright infringement falls on the developers of P2P systems, or the users, will be vital. P2P will also expand from PCs to PDAs and Internet-enabled phones.

One of the issues, which will have to be dealt with in the future, is that of freeloaders - people who consume resources in the P2P network, but do not contribute. According to researchers at Xerox PARC (1), "70% of the users in such systems provide no resources, leaving 1% of the peers to carry 50% of the load." In a free, open system with no obligation to contribute, there will always be users who take and take, and give nothing back. Such concerns, however, may be misplaced - computer files can be repeatedly copied. If I share a file with someone, I don't lose out - the original copy of the file remains on my hard disk, unchanged. In the case of sharing bandwidth over a P2P network, traditional models of resource management don't necessarily apply because bandwidth is replenished over time and unused bandwidth cannot be stored. So unless the system overloads, freeloaders cost other users nothing. There may be a positive externality here; I buy a computer with a lot of processing power, or keep MP3s, for my own benefit. And assuming it costs me nothing to leave my computer on (imagine un-metered net access and cheap electricity), then it costs me nothing for freeloaders to use my resources. See Clay Shirky "In Praise of Freeloaders" (2).

A proposed system for P2P is the use of digital currency to reward stalwart users and guard against freeloading ones. In the case of MojoNation (3), a digital marketplace is created where users who contribute are rewarded with 'mojo' - an internal incentive - while consuming resources uses up 'mojo', so you get what you give. If resources are overloaded, 'mojo' is given for switching to less requested resources. Whether this system works or not remains to be seen. There will always be technological ways of getting something without paying. Such systems also require faith in the currency itself, and a critical mass of users, which MojoNation has yet to achieve. Furthermore, by dissuading freeloaders you could discourage growth and potentially dissuade frequent users.

Another problem to be faced is that of trust: the nefarious elements of the Internet community - spammers, crackers and the like - will target P2P networks. Their open nature makes them easy targets. Anyone can join and if users feel they cannot trust their peers on the network, they will not use it. Even if such concerns are unjustified, the perception that such dangers exist may be enough. Allowing strangers access to your home computer may prove to be a bad idea.

Searching is a definite area in which P2P will prosper in the future. P2P allows queries to be sent to a number of machines. If the initial search is unsuccessful, the query is passed on by each of the original machines to a number of further machines. Thus the search area increases exponentially until a result is found. This eliminates the need for a huge central search engine.

The big guns such as Intel, Microsoft and Xerox have recently stressed the potential of P2P and the likelihood is that P2P will continue to be adopted by big business as a cheaper way of storing and sharing files internally. It is likely that traditional companies will "embrace and enhance" P2P rather than try (in vain) to smother it. The future of P2P may also be determined by fluke, a simple idea may start off small and grow just like Napster did. However, this is now less likely due to the presence of the larger companies who will explore every possible avenue in search for the slightest chance for an advantage over their nearest rivals. Either way, if there is a future for P2P it will be found and exhausted A.S.A.P.

(1) Eytan Adar and Bernardo A. Huberman of Xerox PARC
(2) www.openp2p.com/pub/a/p2p/2000/12/01/shirky_freeloading.html?page=1
(3) www.openp2p.com/pub/a/p2p/2001/01/11/mojo.html

 

Conclusion

Peer-to-peer computing started out of nothing. It became very big very quickly, but once Big Business realised the threat, P2P got knocked on the head. But the genie is out of the bottle and the battle now is for the 'killer app' - the application of P2P that does take advantage of the collective power and content, of the millions of individual PCs worldwide. The application that doesn't fall foul of the legal system, or alienate a market primed by Napster to expect something for nothing. Or more accurately, something for nothing I haven't already paid for.

Various pretenders to Napster's crown have emerged, but as yet none have captured the public imagination. In a way, P2P has retreated to the poorly-lit realm of the nerds. But in another, more accurate way, P2P is quietly being refined and experimented with. It has been adopted by large corporations, and brought under the aegis of the legal system. It has become respectable. Most importantly, the desktop PCs that linked up to give Napster its moment in the sun are still there. They now have even more spare memory, processor power and bandwidth, and hold ever more MP3s. P2P recognises the untapped potential of these peripheral machines. It has to be good for something.