NGI White Paper Research Directions for the Next Generation Internet: Next Generation Network Computing Mike Fisk Technical Staff Member Network Engineering Group (CIC-5) Los Alamos National Laboratory MS B255 Los Alamos, NM 87545 mfisk@lanl.gov 505-667-5119 FAX:665-7793 --------------------------------------------------------------------------- NEXT GENERATION NETWORK COMPUTING Mike Fisk, mfisk@lanl.gov Network Engineering Group, Los Alamos National Laboratory a paper submitted to the Workshop of Research Directions for the Next Generation Internet As realized by the consumer market in the last year, the power of networking is not in its raw communications, but in the applications that it enables. Through networks, people can access information that could not previously be provided to them in cost-effective, current, and useful forms. Part of this added functionality is the ability to combine information from organizationally and geographically diverse sources. As an example, it is now perfectly feasible for a person to order a product over the Internet from a vendor, receive the Federal Express tracking number by e-mail upon shipping, click on the tracking number and be presented with current information from Federal Express about the location of the item, click on that location, and receive a map from the Census Bureau of where the package is. This application doesn't require huge data streams. What it does require is connectivity. The Next Generation Internet project should develop the network of the future that makes data connectivity as accessible and useful as the dial tone. For technology to be useful to the general populus, it will need to be as ubiquitous and easy to use as telephony is. Also exhibited by telephony is the requirement that connectivity must be available to mobile users. For mission critical uses such as defense and healthcare, connectivity cannot be limited to people who are sitting at a desk. The personal computers used today for Internet access are powerful tools for computer specialists, but are too cumbersome for the populus. For the Internet to be used as a communications tool that is as integral a part of life as the telephone, the average user must not be bothered with the configuration and installation of an endless stream of software and hardware. Telephone users would not accept having to reprogram or repurchase all of their home telephones each time a new area-code is created. Instead, changes are made to the infrastructure so that the numbers are automatically accessible to all users. To be useful for health care and national security, the example critical applications mentioned in the NGI, technology must be transparent and dependable, but must also be able to keep pace with the continual and rapid progress made in the fields of computer applications, networking, and hardware. To meet these requirements, new systems must be developed around technology drawn from the worlds of mobile computing, database replication, caching file systems, directory services, public key cryptology, and object technology. To support mobile and disconnected operation well, users must have more than dumb clients connected to central servers. Applications must run locally and often independent from remote systems. However, all new software distribution and configuration should be done automatically. Clients should request all applications from remote servers and store copies locally. Client machines will discover new applications and newer versions of applications periodically. User data must be stored on these servers in a secure fashion that is accessible from any machine that the person may use. In such a manner, networked computers will automatically stay current with state of the art features and users will not have to understand or configure any of the tools that they use to access the Internet. Vendors of commercial software may decide to act as the only network servers for their products. Organizations can provide their own software and may provide local copies of licensed commercial software. Through efficient wide-area directory services, mobile users will be able to locate and access services from any connected location. New techniques in directory services must by developed so that mobile users will be able to locate current versions of replicated and distributed services and applications. Object programming models will aid in the design and distribution of applications that can be easily managed over a network. Current applications are developed for single-user, immotile workstations with lots of local configuration adjustments. Instead, applications should use the network to dynamically locate well-known services and to use universal directory services for user-specific information. The execution of downloaded code is an essential, but risky part of this system. Remote code should in many cases be authenticated and come only from trusted sources. In addition, clients should make use of operating system based compartmentalization and virtual machines to monitor and approve the actions of programs. Further, all information transferred across the network should be strongly authenticated and encrypted. Public key cryptology offers the most scalable forms of encryption and authentication, but the infrastructure of key distribution and trust necessary for wide-area still needs to be developed. Much of the work mentioned above will come as a fundamental paradigm shift for the current computer and software vendors. As such, it will not directly come out of the natural product development cycles of the market in the immediate future. But, by supporting the development of the technologies mentioned above, the NGI initiative will help develop the network of the future that will enable applications that add value to people's lives.