Knowledge management and peer-to-peer computing: making connections

By Eric Woods

“The vision I have for the Web is about anything being potentially connected to anything.” —Tim Berners-Lee, Weaving the Web

Tim Berners-Lee's original concept of an interactive Web is finding realization in the exciting, if confusing and ill-defined, peer-to-peer movement. The best known example of a peer-to-peer application is Napster (napster.com), the infamous MP3 file-sharing application that has given the music industry such a headache. What Napster shares with a diverse set of other P2P applications is that it turns a user's PC from a passive consumer of Web information into an active resource within the Internet environment.

A vision of anything potentially connected to anything is, of course, also a good description of knowledge management. And both peer-to-peer and knowledge management are concerned with connections between people as much as between people and information systems. Peer-to-peer approaches fit well with the emergence of more loosely coupled information-sharing environments such as:

• virtual teams--where remote employees, contract workers or external partners need to share information, but find the corporate IT environment too rigid or controlled for flexible knowledge sharing, and;

• B2B knowledge sharing--where organizations need a managed and flexible environment for information sharing without it affecting their internal IT infrastructures.;

Understanding the implications of such environments is also a central concern of knowledge management. Knowledge management and peer-to-peer enthusiasts share a common desire to realize the true potential of a networked society. Anyone interested in knowledge management--including knowledge management professionals, IT managers and software vendors--needs to be aware of the innovations that peer-to-peer technology is introducing in the areas such as knowledge sharing, information discovery and collaboration.

What is P2P? Driven by the furor around Napster, the concept of P2P brings together under a broad banner a number of projects, initiatives and products that often have limited affinity. There is, therefore, substantial confusion around exactly what qualifies as P2P computing. And some aspects of the peer-to-peer movement have very limited relevance to the knowledge management market—notably distributed computing models such as the seti@home (seti@home.com) project.

In a nutshell, however, P2P computing centers around computers being able to establish a direct connection with each other that enables the sharing of data and system resources with minimal involvement of centrally managed servers. More specifically, in order for an application to be considered as P2P, two primary conditions must be satisfied:

• The application must be designed to accommodate dynamic network connectivity patterns and temporary network addresses (that is to say, peers must be able to form relationships dynamically);

• Peers must be able to operate with a significant level of autonomy (that is with minimal involvement of centralized resources such as a server or, in the case of knowledge management applications, a centralized index or repository).;

While there is incredible hype, confusion and downright silliness being talked about the P2P concept, at its root is a return to the original purpose of the Internet: to allow computers to share information, not just to be passive recipients of Web pages.

P2P and knowledge management

The Internet and the Web took-off because they overcame some of the egregious failures of mainstream corporate IT (notably to provide a flexible, global means of communication and information delivery). Similarly, peer-to-peer's attraction is that it has the potential to circumvent the limitations of current IT—both technical and organizational—in order to improve communication, collaboration and information sharing. In the same way as Napster and similar applications, such as Gnutella, have made it possible for Internet users to share MP3 files across the Net without the need for a central repository of files, so the corporate vision for peer-to-peer is that it will enable users to actively share knowledge and information in new, more flexible ways. Instant messaging is one example of a peer-to-peer application that has found great appeal amongst general users and also has relevance to corporate needs. A more sophisticated example of the sort of services peer-to-peer may offer the enterprise is provided by companies such as Groove Networks , OpenCola and NextPage, for example, which offer a range of communication, collaboration, search and file sharing services.

P2P-based knowledge management solutions will be most attractive where they address problems that are difficult for current software to address. They will not remove the need for centralized solutions, because centralization brings advantages in terms of scope, control and organization. However, there are areas where a decentralized approach has advantages, for example:

• ad hoc or loosely coupled systems--this covers Napster users, but also virtual teams and cross-company B2B projects, where it is difficult to mandate or implement a centrally managed system. This is the prime target for P2P-based collaboration tools, such as those offered by Groove Networks ;

• specialist file sharing--again, while MP3 files may be the best-known example, it is also relevant, for example, to R&D teams sharing working ideas and plans, or the exchange of competitive intelligence among a distributed sales team. NextPage, with its peer-to-peer content management solution, is one of the vendors attacking this opportunity;

• specialist indexing and deployment--the limits of the centralized system represent the limits of the server software—that is, the file types it supports and how it handles them. However, in the P2P environment, a search could be passed between nodes, each of which might specialize in particular information types. If a user, for example, is searching for information on a customer, a node specializing in documents could return recent letters to the customer, but a database node could return relevant sales data. Such a vision is behind the work of the Infrasearch team acquired by Sun Microsystems (sun.com), and now being implemented as part of its JXTA P2P platform.;

P2P knowledge management applications Two specific areas where peer-to-peer approaches are bringing a new perspective to core knowledge management issues are collaboration and knowledge discovery.

P2P and collaboration The problem of supporting the virtual community is central to the developing market for knowledge management technology and supporting services. However, considerable work still needs to be done to turn today's networked environments into spaces for true collaboration. Vendors such as Groove Networks and Endeavors Technology see an opportunity in offering loosely coupled P2P applications for companies deploying intranet- and extranet-based collaborative infrastructure.

Groove Networks, founded by Ray Ozzie, the creator of Lotus Notes, is one of the highest profile P2P vendors. Its core offering is the Groove platform for the development and deployment of collaborative solutions, based on peer-to-peer network communications. The platform consists of the Groove client, underpinned by system, development and integration services.

Groove enables groups of users to communicate and share information using a set of out-of-the-box tools, including instant messaging, live voice, text chat and file sharing, viewing and editing. Information can be accessed from commonly used desktop applications, such as Microsoft Office. Interactions occur with the context of a shared space: a private workplace only accessible to invited users and configured with a set of tools appropriate to the task. All members of a shared space automatically receive changes made by other members while they are connected. Members can continue to work in the shared space when disconnected. Groove saves all changes locally and automatically synchronizes the next time the user connects.

P2P and knowledge discovery

Information discovery and file sharing were among the first established P2P applications. Vendors are now taking those models and extending them into the wider Internet and corporate world. Figure 1 shows how a peer-to-peer infrastructure might support the search process (this is loosely based on the Gnutella model).

OpenCola is one of the companies offering a peer-to-peer model for information discovery that has the potential to compete with existing search technology. With OpenCola Folders, as its product is called, users define areas of interest by selecting relevant documents. OpenCola's “relevance queues” then locate other OpenCola peers and look for documents that match the profile. Relevant documents are returned to the user who can score them for relevance and thus train the Folder about their interests.

Elements of this approach clearly overlap with the services that existing search vendors provide in terms of conceptual search and user profiling. The main difference is that OpenCola does not maintain any central repository or document index. Instead, finding the right documents depends on the resources available within the peer community being searched. This is the strength and the weakness of the peer-to-peer approach. The focus is on communities of peers that can share knowledge on specific areas of interest (in the same way as Napster helps users find music files). There is, therefore, a good chance of documents being of relevance and also of finding less common documents that may not be indexed by Web crawlers. However, it is not a replacement for centrally organized search engines, either in a corporate or Web environment, because there is little control of what may be found (and what will not) and a great dependency on the development of significant support communities. Peer-to-peer knowledge discovery will therefore be a complementary technology to both corporate and Internet-based search engines.

Peer-to-peer and corporate IT: friend or foe? While there is significant interest in peer-to-peer as a concept, it has yet to make a significant impact on the corporate IT world. One reasons is that Tim Berners-Lee's grand vision is the stuff of nightmares for most IT managers. Their systems--and their lives--are complex enough without adding new layers of connectivity and interaction. But IT managers should not dismiss peer-to-peer too lightly.

Of course, it is in the power of IT managers to block (to some extent) use of such application across the corporate firewall. But they will find it difficult to stop it completely if enough users see it as meeting an important need. IT managers need to realize that peer-to-peer will not come knocking at the front door in the guise of a new products from the likes of Microsoft or Oracle, but as tactical solutions to immediate user needs.

This means that IT managers will be fighting a guerrilla war against peer-to-peer--a war fought not on open ground but in the secret byways of personal computing. From the users' point-of-view, of course, it will be a war of liberation to cast off the shackles of the corporate IT, utilize their own knowledge better and put their PCs and laptops to effective use in overcoming the drag of centralized systems.

For IT managers and knowledge management professionals, the best response to peer-to-peer is to understand the frustrations that users have with their current systems, and to be actively working with them to overcome the limitations. And if peer-to-peer offers a feasible solution, they should be in place to guide and monitor experiments, manage the risks and assess the benefits. Be ready to make peer-to-peer a guest rather than an enemy—at least until it breaks the rules of the house.

Eric Woods is research director, Knowledge Management, at Ovum; e-mail exw@ovum.com.