Section 5 describes the key strategy on which IPCRES will be based — to significantly expand the research and development capability in Indiana focused on areas of information technology fundamental to pervasive computing. It is proposed to do this through the establishment of around six world-class research laboratories headed by Distinguished Scientists.
The focus of the research of the IPCRES Laboratories is on research fundamental to pervasive computing. This will be carried out in two broad fields — software technologies and advanced telecommunications. In this Section, five specific research areas in each of these two fields are described in which IPCRES Laboratories could be established. It is not expected that IPCRES Laboratories could be established in all of these areas. However, by specifying this many, it will allow the IPCRES Steering Committee necessary flexibility for the recruitment of Distinguished Scientists and researchers for the Laboratories.
The direction each Laboratory takes may also be guided, in part, by industrial partners or by external funding agencies that are supporting the Laboratory's work. Over the lifetime of any Laboratory, it would be expected that a significant amount of its funding would come from external sources such as the NSF and other federal agencies or by direct support from private industry. This of course is one of the major contributions to economic development that the IPCRES Laboratories will make. In addition, IPCRES Laboratories are not expected to last forever. In some cases, research teams will move the work into the private sector to start new companies. In other cases, the IPCRES Steering Committee will determine new opportunities and new research avenues to be of such importance that a new Laboratory may need to be established.
In Sections 6.1 and 6.2, each of the possible IPCRES research areas is described, its importance explained, the state of the art in it summarized, its grand challenges stated and finally, the potential economic impact of research progress in it is noted. In software technologies these areas are:
In advanced telecommunications these areas are:
The concept of the Information Grid is derived from the analogy of electrical power and transportation grids. We all depend upon the ability of these grids to deliver services, as we need them and when we need them. We don't think about which power generator delivers the electricity to our coffee maker; we are only concerned that the power is available to drive our appliances correctly. As users of each of these "utility grids" we have come to depend on their services: gas and electric power are routinely and reliably available; the roads in our hometown connect to the roads in places where we want to go. We are starting now to think about our access to networked information in the same way. We expect to be able to get a stock quote or a weather report or an airline ticket, all from the Web, any time of the day. We live in the early days of the Information Grid. From our desktop, laptop, or palmtop computer we should be able to draw upon any and all of the computational resources that the Information Grid provides. In the future, we will access the Grid through a type of problem solving portal, a gateway to networked information that consists of software tools that make our interaction with the Grid services seamless and transparent. An IPCRES Information Grid and Portal Laboratory would focus on the definition of these services as well as the portal software we will use to access them.
Why is it Important? One of the most important lessons that we have learned from the development of the Internet is that the strength of its design rests upon a very simple set of core software "services".
These include the software that allows an unreliable network of wires and computers to provide reliable delivery of data, and the ability to locate computers and data anywhere in the world based on simple symbolic names like <www.indiana.edu>. The Internet works because these services are pervasive. They run the same on every computer in the world. Consequently, we can build applications and other more sophisticated services that assume these core services are available and doing their jobs. In this way, systems like the World Wide Web have been developed, depending only upon the core Internet software. On top of the Web, others have built services such as electronic commerce tools that provide additional capability. Each new application stands upon the shoulders of other, more primitive software layers. Grid computing will lay the foundation for the next generation Internet services and applications that will form the backbone of pervasive computing.
State of the Art. The Information Power Grid paradigm for computing is currently based on distributed, high performance facilities interconnected with very fast networks. Currently Grid projects are focused at the highest end of the computing spectrum and involve large national consortia and government laboratories. Computing on the Grid is often thought of as the process of breaking down a computational problem into components that can be distributed across a set of computational resources that offer the most suitable computing equipment, database server, storage server or other specialized devices, for that problem. In addition to hardware, the Grid encompasses software architectures for parallel computing, communications protocols, scheduling, security and policy mechanisms. Current Grid computing applications range from huge computational problems requiring the combined computing power of several supercomputer centers to data mining and analysis to real-time instrument control and data acquisition. What Grid computing applications have in common is the need for an integrated set of services and facilities on which application developers can build. Several approaches to Grid computing are being developed on a national level:
These groups have recently joined together to form the Grid Forum to establish standards for the extended set of Grid services that are required to make this next generation of seamless computing and information possible. Indiana University is working closely with NASAon their IPGproject and with the NCSAAlliance on their system.
Grand Challenges. The current Grid research has focused on the difficult process of implementing the core services of authentication, authorization, privacy, resource co-scheduling and advance reservation that are the needed service elements. However, the grand challenge lies in making all of these services available transparently to the user. More specifically, any Grid access point should be able to easily recognize (authenticate) each user and make all of the resources of the Grid that the user is authorized to use readily available. The user's "information space", i.e. the configuration of files, videos, audio clips and computations that the user last accessed, should be instantly available no matter where or how the user has chosen to access the Grid. Furthermore, access to the Grid would be adapted to the user's current access-point capabilities. If the user is in a 3-D immersive environment, the interaction with the Grid can be very rich. Or, if it is through a cell phone or palm-pilot-like PDA, the access may be limited to spoken English or very simple graphics. While these interaction challenges are really the subject of the IPCRES Human Computer Interaction Laboratory described next, the challenges of designing the protocols that manage the Grid resources and user's information space are fundamental to breakthroughs in this area. In addition, results from an IPCRES Network Agents Research Laboratory, also described below, would be integrated into the core Grid services. The personal agents will be responsible for gathering information and interacting with other Grid services on our behalf while we are "off-Grid".
The greatest grand challenge for Information Grid research is to find ways to design services that can gracefully scale from systems involving small groups of a dozen or so users to networks that will have billions of computers, sensors, instruments, databases and users. The current Internet protocols have proven to be very "scalable," but they may be stretched to the limit. Many feel that they will need to be redesigned in the years ahead. This redesign will be one of the tasks of the researchers in the various IPCRES Advanced Telecommunications Laboratories, but the results will greatly affect the design of such higher level Grid application services as global authentication, personal information space, and software agent management.
Economic Impact. Building an Information Grid with the capabilities described above will provide the software service layer for the next generation of the Internet. This infrastructure, when combined with the advanced network and agent technology described below, will form the foundation of pervasive computing. The total economic impact will be hard to predict. In particular, the Grid infrastructure will support the next generation of electronic commerce, so the impact will be indirect but huge. Grid software lies between the network protocols and the user applications and is called middleware. By itself the middleware business is approximately a $10-billion-dollar-per-year industry. While this is a small part of the software industry as a whole, it is one of the more rapidly growing sectors.
Human-computer interaction (HCI) and human-centered software design research are critical for increased usability and accessibility of computer systems. Pervasive computing and the convergence of telecommunications, data and digital media create new challenges and opportunities for designing systems that are responsive to human needs and that extend human capabilities.
Why is it Important? Access to information is of universal importance, not only to scientists and engineers but for a broad cross section of our society. How we communicate our information needs to computers and how information is presented in a manner consistent with our immediate interests will ultimately determine whether we drown in information or use it profitably. In addition to information discovery and presentation a pervasive computing infrastructure can also mediate human communication and collaboration. The recent development of global high speed research networks has been driven by the emergence of global-scale research consortia; the need for communication through networks is a central requirement for support of basic science, engineering and other disciplines. The private sector has a similar need for high speed networks and collaboration tools to bring together design and engineering teams from around the world for product development. The convergence of computing, telecommunications and advanced display technology is creating a rich environment in which new modes of communication can emerge. Together these general areas resolve into six key research fields for an IPCRES HCI Laboratory: advanced interface methodologies, computational linguistics (and automatic translation of human languages), collaboration technologies, telepresence, augmented reality, and remote operation.
State of the Art. Speech input and output are used now in "hands-free" applications such as automobile navigation systems and airplane and fighter aircraft cockpits. Windows, icons, mice, pull down menu (WIMP) interfaces, e.g. the Macintosh Finder and Microsoft Windows, are being re-interpreted in personal digital assistants and handheld computers to include handwriting and speech recognition.
Semiconductor manufacturers such as Philips are developing chip sets for wireless communication to link mobile handheld computers to networks allowing access to information and people any time and anywhere.
Research projects at MIT in language understanding and speech synthesis as query interfaces to information sources have demonstrated the usefulness of this approach. The DIPLOMAT project at Carnegie Mellon University is successfully translating several human languages in real time, including Serbo-Croatian, Spanish, Korean and Haitian Creole. The potential of these projects for improving global, cross cultural communication is obvious.
In addition to universal access to information, pervasive computing also promises to create new ways for people to interact, and the user interface is where that interaction takes place. New interaction modes broadly grouped under the heading "collaboration technologies" include digital video and audio conferencing, portals, shared virtual spaces, and tele-immersion. Portals (described elsewhere in this document) are personalized collections of tools and data that provide a window on some field or activity, collecting all relevant resources in one interface. Shared virtual spaces are portals extended to include several simultaneous participants. Tele-immersive shared virtual spaces use virtual reality technology to project a 3-D space in which several people can collaborate in 3-D design and evaluation, or explore data sets represented in 3-D space.
Tele-immersion (T-I) joins together high performance computing, high speed networking, and virtual reality (VR) hardware and software to allow users in different locations to collaborate in a shared 3-D virtual environment as if they were in the same room. Tele-immersive applications combine virtual worlds, 3-D models, audio, video, simulations, and many other technologies to create a shared workspace in which users can meet to consult, evaluate designs, plan and review simulation results. Some T-I applications include: shared 3-D drafting and engineering spaces; creating and interacting with immersive 3-D representations of data; tools to combine, explore, analyze, store, and edit data sets; and data mining of large data sets and real-time sensor data streams. Tele-immersion has become a subject of research worldwide and has entered the lexicon of the Next Generation Internet (NGI) program and Internet2 project as well as international networking programs.
Use of virtual reality technology to overlay 3-D information on the real world is called augmented reality. Augmented reality can be used to display tactical information on a fighter pilot's heads-up display or MRI data on a surgeon's view of a patient. Physical plant workers could "see through walls" to locate electrical wires or plumbing quickly and without using complex multi-layer diagrams. Despite the current cost of head-mounted displays many applications for augmented reality in maintenance, repair and information display are currently being developed. The combination of "just in time" information displayed in a spatial real-world context holds promise as a powerful human-computer interface technology.
Related to tele-immersion and augmented reality is tele-operation, or the use of VR equipment to permit the operation of robotic devices from a distance. Typical applications include exploration of hazardous or inaccessible environments such as the deep ocean, inside a nuclear reactor, or the surface of Mars. A recent spectacular example was the use of a robotic submarine to explore the wreck of the HMS Titanic.
Recent advances in human-computer interaction are very promising. Speech recognition and input is a modest commercial success and promises to become more important as this capability is added to "hands-free" environments like cars and airplanes. Text to speech is an important way for disabled persons to participate in computing and the promise of the Web. Machine vision for gesture recognition has been demonstrated, and haptic (force feedback) devices are available for research and entertainment applications. Distributed computing, instruments, mass storage facilities and new interface technologies will allow groups of people separated by space and time to work together in shared virtual workspaces.
Applications for advanced interface technology include:
Grand Challenges. The primary challenge in human-computer interaction is to understand and use human modes of communication such as speech, handwriting, gesture, touch and stereoscopic 3-D vision to communicate more naturally with information systems.
Speech input and output with natural language understanding, as well as vision and gestural interfaces, are more appropriate than current WIMP interfaces for many tasks that humans must perform with computers. Research in advanced interfaces seeks to apply an understanding of human communication in multiple modalities (speech, text, image, video, gesture, facial expression, handwriting, etc.) to improve human-computer interaction, both qualitatively and quantitatively. The problem of automatic processing of multimodal (speech and gesture) human communications, including applied computational linguistics for language understanding, discourse in a variety of tasks, and framing of discourse in a format appropriate for available interface hardware, and machine translation is a major challenge. Gestural input combined with machine vision provides a powerful way to bring computers into our 3-D world and to allow them to understand our immediate intentions for objects in the field of view. Creating a context for human-computer interaction and using appropriate interface modalities in context is a key problem to be solved in support of collaboration systems.
Challenges in display technologies, networking infrastructure, and 3-D navigation abound in developing next-generation tele-immersion systems and applications. Some of the more pressing problems are developing lower-cost 3-D graphics hardware and projectors, dealing with groups of related and often synchronized network flows, and navigation issues in shared virtual spaces (leading, hinting, following, and recording).
Economic Impact. The fundamental impact of advanced interfaces will be the disappearance of computers as we think of them now. "Ubiquitous" systems will become "pervasive" in that we will think of them less as individual machines and more as links to a larger information space. Incorporating human-like modes of interaction such as natural language understanding, speech, and vision will make information systems more accessible and usable. Personal computing devices so equipped will help us communicate with each other in richer ways and will make it easier to organize and use information more productively. The promise of advanced interfaces to improve collaboration at a distance is great and new interfaces such as real-time machine translation of spoken and written language have obvious applications in a global economy.
The quality of teamwork in design, engineering and maintenance tasks can be greatly enhanced through interfaces based on VR and other advanced display technologies. Tele-immersion will bring geographically dispersed groups together in shared virtual spaces; augmented reality overlays information such as 3-D CAD drawings on the real world, and tele-operation of remote robotic devices will provide "better than being there" access to remote or dangerous environments.
Applications for tele-immersion can be found in a broad range of fields where it is an advantage for teams to share large data sets or 3-D models, or work together at a distance. These include data analysis in science and engineering, interactive instrument control, virtual prototyping, usability testing, medical imaging and market testing. Several recent tele-immersion projects point to the necessity and viability of this research program. Caterpillar, Inc. design teams have demonstrated the use of shared virtual spaces in product development, using shared CAVE systems in the US and Britain. General Motors is using T-I systems for evaluation of the layouts of automobile interiors. Architectural firms are increasingly using VR and T-I techniques for client walkthroughs. Scientific applications of T-I include a shared space for data-mining the structures of molecules determined by X-ray crystallography and the evaluation of hydrological and ecological data from the Chesapeake Bay. Education and training applications include use by NASA to train flight and ground crews before the Hubble repair mission, medical training in ENT anatomy and electrocardiograms, and many others.
The term Smart Device refers to a device that is connected to the network and can respond to requests for its specialized services from remote clients. Examples include communication devices like pagers, cell phones, and small sensors such as thermometers, accelerometers, audio and video cameras and bio-sensors of various types. Larger examples include telescopes, wind tunnels and MRI, CAT and other imaging systems. Smart devices also include the host of tools and appliances that contain embedded computer systems. This last category is rapidly becoming ubiquitous in our home and workplace environments. It includes automobiles, factory machines, new home appliances, entertainment devices and, of course, general office equipment. The embedded computers in these systems are very small: they may be woven into fabric or, someday, painted on surfaces. The purpose of these embedded computers is to provide whatever programmability the host device requires and to act as the communications window between the device and outside clients. Pervasive computing happens when these small systems begin to communicate and inter-operate over a common network.
Why is it Important? In many ways it is the smart devices that distinguish the age of pervasive computing from what we see around us now. These systems represent the contact point between the analog, mechanical, biological world in which we live and work and the digital information space that is evolving around us. These smart devices are already driving massive changes in manufacturing: everything from quality control and robotic assembly to plant safety is becoming dependent upon this technology. Our transportation systems, communication systems, and health support systems are, or soon will be, transformed by these devices.
State of the Art. The software that drives smart devices is not unlike the software in our conventional computers; the primary difference is size and speed. Because embedded computers are so small, they can't run standard operating systems like Windows or Unix. Instead they run very light-weight operating systems such as stripped-down, highly simplified versions of Sun's Java "virtual machine" or other, very simple proprietary systems. In the larger devices such as PDAs like the Palm-Pilot, more complete operating systems can be run. These include operating systems like PalmOS and Windows CE. The one thing that all these systems will eventually have in common is a way to talk to the network through either a direct connection or a wireless connection. In many cases the device itself will not be able to handle the complexity of the TCP/IP protocols of the Internet. In these cases the devices might communicate to a software agent running on a conventional Internet host that can speak the device's more primitive language and provide the protocol translation and act as the proxy for the device in the external digital world. The design of these embedded software tools and agents is already emerging as a substantial part of the software industry. Sun's recently introduced Jini system is devoted entirely to this market and Microsoft and others are working on different solutions. The state of the art is still very fluid.
Grand Challenges. There are three initial challenges in the design of the software infrastructure for smart devices. The first involves the ability of a device to announce its presence to the rest of the information sphere as soon as it is turned on or connected to the network. If the device is extremely small or limited, a proxy must be established for it on a network host. The second challenge is in finding ways for other devices or Internet clients to interrogate the device to understand its capabilities. Third, one must have a standard protocol for the smart devices to announce changes of state or condition to interested clients. Taken one at a time, these are not difficult technical problems and partial solutions exist. However the greatest single challenge for pervasive computing is to manage scale. That is, how does one build a system in which all of these software services work seamlessly when the number of smart devices on the network reaches into the billions? Given that these smart devices will "know" a great deal about us as individuals, how do we secure this knowledge so that our privacy is respected? Finally, how can a distributed network of smart devices be configured into intelligent applications?
Economic Impact. Network-enabled manufactured goods will become ubiquitous and the software industry to support them can be expected to grow very rapidly over the next ten years. As the basic software standards emerge, we can expect to see a secondary market of software applications grow along with it. These applications will allow us to orchestrate the smart devices we control and they will make the type of autonomous decisions needed to maintain the device network. For example, these applications could notify us when our car requires service and pre-order any parts that may be needed. The same type of application can advise us when we are leaving the office about the contents of our home refrigerator, so that the necessary supplies can be acquired on the trip home. The same application may be given the authority to re-order supplies automatically and make sure that they are delivered to our home in a timely manner. However the greatest impact of this technology will be in the workplace. These pervasive computing technologies will define a new revolution in the efficiency and productivity of our factories and places of work.
Nicholas Negroponte, of the MIT Media Lab, wrote, "the future of computing will be 100% driven by delegating to, rather than manipulating computers". To accomplish such functionality, future software must be autonomous, reactive, proactive, and social. Autonomy means the software can act without direct intervention of humans based on imprecise input. Responsive software can perceive the environment, for example the Internet, and respond in a timely fashion to changes that occur. Proactive means the software can take an active role to pursue a goal-directed behavior and take initiatives when needed without being told what to do next. Being social implies that software can interact with other software and with humans in order to complete a task in which it has engaged. Software programs that encapsulate these properties are called software agents. A laboratory devoted to the development of multi-agent systems and their applications will be an important component in the pervasive computing initiative.
Why is it Important? Pervasive computing refers to a ubiquitous fabric of intelligent instruments, information sources and information analysis tools tied together by wired and wireless networks. Intelligent software agents will be an important vehicle through which users can "delegate" tasks on their behalf. The information grid of the future will have to deal with terabytes of multimedia data held in heterogeneous databases and billions of requests for information each day. Multi-agent systems that discover, monitor, and filter information will be necessary for users to sort out vast amount of information available on the Information Grid.
State of the Art. Agents are the next major computing paradigm and will be pervasive in the future market. Indeed, the potential use of this new software paradigm is explored in many areas ranging from industrial applications (e.g., process control, manufacturing, air traffic control), commercial applications (e.g., information management, electronic commerce, business process management), medical applications (e.g., patient monitoring, health care), and entertainment (e.g., games, interactive theater and cinema). High-speed networks that would allow the agents to roam and collect information are becoming a reality. Small computing devices ranging from hand-held PDAs to wearable computers will be common in the immediate future. Given these hardware infrastructures, agent software technology will be the intelligent software solution that will make ubiquitous computing possible.
Grand Challenges. Software agents are still at their infancy. Ideally, the task of a software agent is to fulfill a task for its user. In order to accomplish a task, the agent may need to use the Information Grid to access new information, may collaborate with other agents to find out if they have solved similar problems before, and if so, negotiate to acquire the knowledge to solve the problem, and finally deliver the result to the user's environment. Sometimes the knowledge to solve a problem may already exist on a legacy system. In such cases, an "agent wrapper" may be necessary to obtain the information. Additionally, agents will have to deal with managing text, semi-structured information, multimedia data, and translating between representations. In addition to discovering and monitoring information resources, agents have to assume the task of analyzing, summarizing, integrating, and fusing information from heterogeneous sources and present the information in a meaningful way to the user. The user's environment may change from time to time from a desktop computer to a hand-held PDA. The agent must recognize such environment changes and adapt to interacting with the user.
The above stated challenges are related to agent-based application development. Even more important, there exist no standards to address the pragmatic and engineering aspects of agent-based system design, and no production-quality software support for building agent applications, and such an infrastructure is definitely needed to develop future agent-based systems.
Economic Impact. Future growth in agent-based software systems will be in two primary areas: agent design tools and agent-based applications. Significant software development activities will happen in these two aspects. Similar to numerous Web-design tools that exist today, there will be a critical need for software tools that would allow one to design an agent application. Similarly, agent-based application development will make a significant impact on the software industry. The next generation Web will be an agent-based portal of the user. The proposed IPCRES Laboratory in this area along with the "open software" concept will provide an excellent infrastructure for agent-based system design and development. Many existing and future software development companies will be attracted to this facility.
The term Open Source Software refers to the development of a software system where the original computer source code is developed and made available in the public domain, as a result of which the software is developed and maintained by teams of contributing authors. A second form of Open Software refers to the design of standard protocols that become widely used and are essential for applications and devices to communicate and share information. A laboratory devoted to open software would help to coordinate the efforts of this community authorship process and would work with private companies who incorporate components of open software into commercial products. Such a laboratory could provide support to companies or institutions that use the open software or protocols, and could provide a supporting venue for meetings and conferences about the future of specific open- source software projects and a well managed repository for keeping the current versions of the software available to the public.
Why is it important? The promise of pervasive computing rests upon ubiquitous standard protocols and software that drive the network and telecommunication infrastructure. Without complete consistency in the way Information Grid services are established, implemented and maintained, applications and devices that rely upon the software middleware will not operate. The disjointed and uncoordinated development of television and cellular telephone technology, neither of which inter-operates smoothly across international boundaries, is a good example of what can go wrong if international open standards are not in place. The Internet protocols, while not perfect, stand as excellent examples of standards that have evolved over time by the efforts of a public, open process. The next generation of the Internet cannot happen without this process continuing.
State of the Art. The Internet standards activity for TCP and IP and many other protocols are well known community processes; much of this activity is conducted through the world-wide volunteer efforts of the Internet Engineering Task Force (IETF). The World Wide Web is also based on open standards and software developed through an international collaboration of the World Wide Web Consortium (W3C). Netscape, Inc. has placed the source code for its famous Web browser in the public domain and it is now extended and maintained by the open software process. The Web server that is most commonly used is the open-source Apache server. Other new Web technologies such as XML and WebDAV, which will greatly enhance the current generation Internet, are the results of community processes and have received enthusiastic support from companies like Microsoft, Sun and IBM. The only serious alternative to Microsoft Windows is the Linux Open Source operating system. Linux as a public domain product has become so important that IBM and many other companies now develop versions of their applications to run on it. An entire industry has sprung up around supporting Linux in commercial environments.
Grand Challenges. There are several major challenges facing the open-source software market. The most important is to provide continuity of these public domain projects. It is easy for a highly successful open-source project to splinter into semi-proprietary products, each with its own small differences. As these differences accumulate, the benefit to the community diminishes. The importance of establishing a IPCRES Laboratory that can maintain a non-profit center of gravity for some of these projects cannot be underestimated. In the case of open standards like those of the Internet protocols the IETF maintains a solid foundation. But other important open software products that may emerge from IPCRES and other research centers will need a home that can provide support and coordination of their maintenance and evolution.
Economic Impact. While it may be hard to imagine that anybody can make money from free software, the reality is that an entire industry has sprung up around it. Companies like Red Hat Software (and at least three others) are distributing packaged versions of Linux to customers who would prefer to purchase software from a reliable source that can provide service. In August 1999, Red Hat staged an extremely successful IPO and others are expected to follow soon. In addition to this software packaging and support industry, many vendors like IBM, Silicon Graphics, Hewlett-Packard, Dell and others have or are planning to ship products or systems that run Linux out-of-the-box. This is not just a reaction to the near monopoly status of Microsoft Windows. Linux and the other open-source Internet applications and services have become too important for these companies to ignore. An IPCRES Laboratory for open-source software and standards can be a unique non-profit center, around which profit-making ventures can gather. Today, some of the best programmers around the world spend time contributing to and working with open-source software. An IPCRES Laboratory in this area will act as a magnet to attract many of these people to Indiana.
Advanced network-based applications such as remote instrument operation, converged telecommunications and grid-based computing have special requirements for network performance that current Internet architectures cannot deliver. Such applications need to be able to request assurances from the network for specific, guaranteed, end-to-end transmission parameters including bandwidth, maximum packet loss, transmission delay and variation in the delay. The current Internet architecture is characterized as best effort, meaning the network provides no guarantees regarding performance but simply does the best that it can. Best-effort networks can't meet the needs of advanced applications. Networks that can provide quality-of-service (QoS) assurances must be developed.
Advanced applications can have enormous requirements for bandwidth. Visualization and grid computational systems require large amounts of bandwidth. As human-to-machine interactions become more sophisticated, including speech and visual interactions, bandwidth demands will increase. Converged telecommunications including voice and video will also drive demands for bandwidth. The development of services for high bandwidth communications is multifaceted. The underlying circuits, typically fiber optics, must provide carrying capacity, the network nodes which route traffic from source to destination must have the capacity to switch large numbers of data packets, and host computer software must be able to support the high performance connection.
An IPCRES High Performance Networking Laboratory would perform basic research into network support for advanced applications. In collaboration with the Internet research community the Laboratory will work for the specification and development of new Internet architectures and protocols for QoS. The Laboratory will perform research into network service implementations and develop specifications for host implementations of advanced network service protocols. The Laboratory will work closely with commercial organizations for the development of high-capacity fiber optic and network router technologies and will deploy large-scale testbed networks in support of technology development.
Why is it Important? Well over 90% of today's network-based applications run adequately on "best effort" networks. These applications, such as email, Web browsing and file transfers don't require extended, real-time, network-based interaction or conversation. But underlying pervasive computing is a shift to advanced network-based conversations that are synchronous, multilateral, have real-time dependencies and in some cases require significant bandwidth. The applications of pervasive computing will have strict requirements for bandwidth and network QoS. Networked devices, sensors, instruments and interfaces will be synchronized on real-time tasks that can't wait on network congestion. In order to achieve their full computational potential, grid-based computers can't wait on the network for interchange of partial results. Packet loss in converged telecommunications networks will lead to unintelligible conversations. Remote operations of instruments require strict bounds on the quality of the underlying network. The success of pervasive computing will rely on networks that can provide high bandwidth and quality-of-service assurances.
State of the Art. Current commercial Internet and high performance research and education networks (HPRENs) such as the National Science Foundation-sponsored vBNS, the Internet2 Abilene network, NASA's Research and Education Network (NREN), the DoE Energy Sciences Network (ESnet), the Asia-Pacific Advanced Network (APAN) and Canada's HPREN CA*net2 are best-effort-based Internet Protocol networks. Significant work is underway in IETF working groups to define protocol and architecture standards for implementing quality of service mechanisms within best-effort networks. The leading architecture under development for end-to-end network QoS is called Differentiated Services (DiffServ). The DiffServ architecture specifies two components, the underlying network protocols that control handling of data packets according to QoS specifications and Bandwidth Brokers. Bandwidth Brokers are the agents that process requests for QoS, perform the authentication, authorization and accounting, and arbitrate inter-network QoS.
The development of high bandwidth network technologies is being led by the development of terabit-speed routers and the development of Dense Wave Division Multiplexing (DWDM) optical technologies. In order to achieve higher levels of router performance much of the logical function is being moved from software running on general-purpose microprocessors to custom-built Application-Specific Integrated Circuits (ASICs). DWDM optical transmission techniques permit multiple signals to be transmitted on a single optical fiber using different wavelengths of light.
Grand Challenges. Although QoS architectures and protocols are under development, little work has been completed in the realm of implementation and management services. The types and priorities of applications and required resource guarantees must be defined. Management policy and techniques for distributed governance must be developed. Resource allocation mechanisms across multiple domains, bandwidth brokers and policy servers must be developed. Baseline standards, reference configurations and dependability parameters for auditing performance to contract, measurement tools and common appropriate response to non-compliance and priority conflicts in inter-networks must be developed. Second order effects must be investigated and understood. QoS specifications for IP Multicast services (one-to-many transmission) must be developed.
The Grand Challenge facing the development of high-performance network technologies lies in the development of optical inter-networking and the integration of DWDM with terabit routers. In optical inter-networks the data are directly placed onto DWDM fiber networks without the use of telecom transmission gear based on the Synchronous Optical Network (SONET) technology. When employed with optical cross-connects of the DWDM wavelengths, routers will be able to direct packets on networks that are logically constructed from wavelengths of light.
Economic Impact. An IPCRES High Performance Networking Laboratory will complement the existing Indiana University initiatives in high performance networking such as the Internet2 Abilene Network Operations Center, the TransPAC initiative and other initiatives. This will strengthen the role of Indiana University and the State as an intellectual center for high performance networking. In the short term, significant relationships with leading network equipment vendors and the Internet research community will be developed. Long term, the Laboratory should become a magnet center for co-location of commercial research and development facilities and for development of start-up technology companies. The chief economic impact of the Laboratory will be a second order effect, in that high performance networking is a fundamental enabling technology for the gamut of pervasive computing.
In simplest terms, wireless networking can be characterized as two types of technologies. One technology is used to construct links between the nodes of network backbones, such as point-to-point links between buildings on a campus, or incipient high-speed wireless Internet service provider connections to residences. The second form of wireless technology directly enables devices and information appliances to communicate in local and global spheres. While both are important, this latter technology is relevant to pervasive computing in a very fundamental way.
Wireless networking support for pervasive computing will not be provided by a single technological solution. Ideally, the multiple technologies will rely on a common standard Internet Protocol (IP) for internetworking, but distinct physical transport technologies will be required to satisfy the requirements of the unique domains.
Many smart devices will operate across several of the network technologies. Each of the network layers will have interfaces to other layers enabling intercommunications of devices and reach to global scope. An IPCRES Wireless and Mobility Laboratory will perform basic research into the broad scope of network requirements for pervasive computing and will translate the research into the development of open standard protocols, implementations, testbeds and management techniques.
Why is it Important? Wireless networks will be one of the fundamental enablers of pervasive computing. As described in Section 3, pervasive computing refers to a ubiquitous fabric of intelligent instruments, appliances, devices, sensors, and computational and information resources and tools. The thread of the fabric is communications. Wireless networking is the technological underpinning that enables pervasiveness.
Pervasive computing will not be served solely by wireless networking. Wired networks will continue to play an important role especially in backbone and access networks, but a plethora of in-situ smart devices and appliances cannot simply be connected by more and more wires. Of even greater importance, a key component of pervasiveness is mobility, which can only be served by wireless networking.
State of the Art. New advances have been made in the last few years in the realms of device, personal and local area networking. The Bluetooth specification, released in July '99 provides an open-standard platform for low-cost, short-range, medium-speed interconnection of small groups of intelligent devices. Bluetooth-enabled devices are expected to be available in the year 2000. The IEEE 802.11 specification for wireless local area networks supports speeds up to 2 megabits per second and coverage extending 100 to 500 feet. Peer-to-peer and central access point topologies are supported. Wireless wide area networking is supported by traditional call-based technologies such as cellular phone and by recent developments of simple message passing services such as AT&T's Wireless IP and BellSouth's Intelligent Wireless Network.
Grand Challenges. The current wireless and mobile technologies have not been developed with the requirements of advanced pervasive computing in mind. Bluetooth was a step in the right direction, but falls short of the scale and speeds required. Specifications will need to be built from the ground up, including requirements for greater numbers of peers inter-operating at higher speeds, supporting converged applications. Global always-on networking will be required for pervasive computing. Smart devices will be actively conversing or listening on the network at all times. Call-based technologies and economic models must be replaced. Mobile identity must be developed; the network must recognize and find users and devices wherever they are. Techniques and protocols for instantaneous network routing changes must developed to support the roaming user. Techniques for management of the networks must be understood and developed. Wireless networking for pervasive computing is an extraordinarily complex technology and involves huge segments of the computing and telecommunications industries. The politics of advancing an agenda in such a competitive industry segment will be just as complex as the technology.
Economic Impact. An IPCRES Wireless and Mobility Laboratory will develop significant relationships with the leading computing and telecommunications corporations and with the Internet research community. The Laboratory will become a leading global intellectual resource for wireless networking in support of pervasive computing. The relationships and intellectual leadership will lead to the development of commercial and consultative ventures centered on the Laboratory. Wireless networking for pervasive computing will be the next telecommunications revolution.
Convergence is the integration of what are now the discrete modalities of voice, video and data communications and applications into a single, coherent network and applications framework. In its simplest, most commonly understood form, convergence is considered to be the development of networks, including new architectures and protocols, that support delivery of the discrete communications modalities on a single IP network. This aspect of convergence will significantly reduce the costs of infrastructure and human resources, simplify management of network services, and provide a rich environment for deploying a vast array of new services. In its fullest sense, convergence, relying on underlying support of the network, extends to the development of new applications and services that combine the powerful elements of computing, media and communications.
An IPCRES Convergence Laboratory will perform research into the network requirements of converged applications and services and will perform research and development of open software frameworks that support the development of converged applications and services. The Laboratory will look broadly at the opportunities of convergence and will develop prototypes and reference implementations of applications and services. The Laboratory will serve as an industry resource for expert consultation and dissemination of information on convergence technologies.
Why is it Important? Human cognition involves the simultaneous integration of multi-modal forms of communication and sensing. The discreteness of the technologies of voice, video and data communications and applications places artificial constraints on the effectiveness and richness of technology-mediated communications. The discreteness is purely an artifact of the state of the art in current technology and the application of technology. Progress in the development of human-to-computer interaction and technology-mediated, human-to-human interaction will be keyed to the convergence of computing, media, communications and the underlying network infrastructure.
In addition to a revolution in technology-mediated communications, the convergence of computing, media and networks will lead to a plethora of as-yet-unimagined applications and services in consumer and business markets, and in education and research.
Pervasive computing will require seamless technology-mediated interactions via multi-model forms of communications. Convergence of computing, media and networks is one of the building blocks for the development of pervasive computing applications.
State of the Art. Networks today support combined voice, video and data transmission via discrete channels carved out of a common transport. True converged networks are developing, but will depend on the development of new protocols and services, and in particular, high performance networking technologies such as network quality of service (QoS) (as described in Section 6.2.1). Some rudimentary converged applications are developing today, but the field is in its infancy. A good example is the development of cellular phones that integrate email and simple text-based Web browsing. Conspicuously absent is an open and standardized software framework that serves as a building block for the development of new services and applications.
Grand Challenges. Much of the progress in convergence will rely upon the development of converged high performance networking, including formulation of high bandwidth services and network quality of service. A challenge for an IPCRES Convergence Laboratory will be to perform the basic research that will lead to a complete description and understanding of the network service requirements of the broad scope of converged applications. This knowledge will guide the development of open standard network protocols and network services.
Converged applications in pervasive computing will require standardized open software frameworks and components, and application programming and communications interfaces. A complete and extensive understanding of the broad scope and potential of converged applications must be developed, which will serve as a guide for the development of standard frameworks, components, interfaces and protocols.
Economic Impact. The development of converged network services has the potential for profound economic impact. Converged networks can significantly reduce costs of infrastructure and human resources and simplify management. An IPCRES Convergence Laboratory will be at the center of this telecommunications revolution. The Laboratory will serve as a magnet center for colocation of commercial research and development facilities and for development of start-up technology companies. This Laboratory will serve an important consultative role for the industries and organizations in the State and will certainly foster the development of linked consulting agencies.
The second tier of convergence, that of computing, media and networks, presents even greater opportunities. In this area entirely new industries and services will develop, and many of the big new ideas will come from small, dynamic technology-savvy entrepreneurs. An IPCRES Convergence Laboratory will serve to foster a local entrepreneurial environment and serve as a continuing intellectual center for the development of convergence technologies.
Pervasive computing is neither possible nor is it even desirable without a strong privacy and security framework that must first be implemented on top of the open network environment. In order for pervasive computing applications to be widespread and successful, any group of network principals should be able to exchange data of any kind, including executable code, without the risk of that data becoming captured, corrupted, or substituted by an unauthorized party, including network administrators. The network principals that participate in those transactions must themselves be robust and incorruptible so that they can deal with any piece of active code or data received over the network without a risk that their own functions and locally maintained data could be compromised.
Why is it Important? It is fairly well known how to accomplish this level of security in the world of UNIX and mainframe computing. DCE, for example, provides a comprehensive environment that can be used to couple a variety of computer systems and devices within an enterprise and beyond. On the other hand recurrent epidemics of computer viruses that on occasion have brought down whole email systems at numerous institutions and that plague the world of personal computers attest to generally flawed and careless operating system and application level designs in that domain. If by pervasive computing we mean that even more lightweight entities are to be attached to the network, whereas at the same time even more of our everyday activities are to be carried on over the network, there is the potential for chaos that would render the network inoperable and useless for any serious activity. Hence the development and establishment of widely agreed upon protocols, methods, and frameworks that would guarantee privacy and security of network principals and network transactions are a condition sine qua non of pervasive computing ever becoming a reality.
State of the Art. Among the many reasons why Java attracted so much industry attention is that Java encapsulates a quintessence of distributed computing mechanisms developed in the world of UNIX in a relatively small and lightweight package that can be easily and cheaply deployed on small network attached devices or within applications that are used to communicate with the network (for example, Web browsers). This replacement of a fairly heavy and complex UNIX framework with the Java Virtual Machine is what makes pervasive computing possible. Java provides a broad range of security mechanisms beginning with an in-built theorem prover that inspects a downloaded byte code on the fly, through digital signatures, message digests, public encryption key algorithms and support for X.509 certificates. The state of the art algorithms and methodologies employed by Java are known from popular stand-alone tools such as PGP (e.g., NIST DSA, MD5, and SHA) and larger systems such as DCE (e.g., DES and Triple-DES).
Grand Challenges. But Java on its own currently lacks the rich framework that would comprise trusted public directories and services that could issue certificates to network principals. Java lacks what in the DCE world, for example, is provided by Cell Directory Services and by the Registry, what in the NIS+ world is provided by system tables, what in the OSI world is provided by the X.500 model for Directory Services, and what in the IETF world is provided by LDAP (as specified by RFCs 1777 and 2251). There is no persistent storage management and no transaction engines. Consequently enterprise-wide applications developed in Java Beans have to couple to non-Java external service providers, for example, DCE, CORBA-services, IBM TX-Series, M3, LDAP, Oracle, Sybase, or even DCOM. Most of those have Java interfaces; nevertheless the current landscape is that of a hodge-podge of hastily wired up solutions whose maintenance is costly and cumbersome and longevity questionable, and which, in most cases, cannot scale beyond the enterprise level to encompass the whole network.
An example of a service that does scale and that is universally accepted is the Domain Name Service, DNS. The other technology that is perhaps less mature, but, nevertheless, broadly accepted is that of LDAP. Currently developed systems for meta-computing such as Globus make extensive use of LDAP to manage distributed supercomputing environments. LDAP version 3 can be integrated with MIT Kerberos 5 so that LDAP entries are protected against unauthorized access. Solutions such as DNS, LDAP, and Kerberos are extremely robust and proved their value many times in the open and highly demanding Internet environment. In combination with Java and enhanced by the public encryption key technologies utilized in PGP and Secure Shell (ssh) they may eventually lead to the establishment of open, trusted, secure, and universal directory and certificate services that would become a foundation for the development of lightweight pervasive computing. The challenge is to make this happen in cooperation with the industry, academia, and with the open network community, and to develop genuinely useful demonstration systems that would illustrate both the merits and the how-to of a well integrated pervasive computing architecture.
There are particular problems associated with the security of high performance networks. This would provide the opportunity to establish an IPCRES Laboratory in this area. In doing so, collaboration would be explored with the CERIAS Center at Purdue to which the Lilly Endowment gave major funding in 1998. Discussions have already been held with CERIAS about this possibility.
Economic Impact. The deployment of this framework will have major economic consequences. Although the tenets of "e-commerce" are loudly touted nowadays, real "e-commerce" is impossible without security and privacy protecting the buyer and the seller. It is impossible without a framework for authentication and authorization of network principals, without a framework protecting confidentiality of stored data and data transactions, without techniques to prevent repudiation of contracts, without auditing. The framework must be widely deployed and universally accepted. It must be lightweight, so that corresponding processes can operate without hindrance within hand-held devices, within small network attached devices, within network attached sensors and actuators. And it must be scalable, so that hundreds of millions of certificate requests and directory lookups generated by all those devices constituting pervasive computing would not bring network services down.
Stripped of its paraphernalia, all computing, including pervasive computing, is about data — generating, transforming, manipulating, and storing data. In a traditional computing context data are stored on local devices (e.g., disk drives or tapes), while in most distributed computing environments (e.g., DFS, NFS, AFS, Novell), data are no longer local but instead are served to the network from shared devices. But unlike these more traditional distributed computing environments, which typically have a limited number of specialized data servers, in the pervasive computing environment every network principal is potentially a data server. For example, a camera that is directly attached to the network serves images; an activator may simultaneously be a sensor that captures response data that are transmitted back to the network. Whereas some of that data may be transient, some may have to be stored for a long time: days, months, years, tens of years even. And whereas some data may be distributed liberally, other data may be highly confidential. As pervasive computing will ultimately encompass billions of devices sharing network, storage, and CPU resources, systems and techniques will have to be developed to provide secure high bandwidth access to data and to ensure security and integrity of all data transactions.
Why is it Important? A small, lightweight, hand-held device that provides network access in the future is unlikely to contain moving parts such as disk drives and tape drives. Rather we should think of a device like that more in terms of an intelligent telephone, perhaps with a miniature keyboard or a pen input, a small high resolution liquid crystal display, and possibly additional miniature input and output devices, e.g., cameras, microphones, and speakers perhaps a device that resembles a palm-top computer. That device will store a certain amount of information in non-volatile memory, but ultimately it will have to go out to the network and connect to distributed persistent storage devices in order to obtain access to large amounts of data. It is indeed preferable that valuable information is not stored on a device like that, because it is vulnerable to theft, breakage, and data corruption. Many other devices that constitute the fabric of pervasive computing will likewise interact with large remote data stores. In order to make pervasive computing possible, data must become pervasive as well.
State of the Art. As is the case with security and privacy, it is fairly well known how to deliver distributed data services to clients in the world of mainframe and UNIX computing. The combination of High Performance Storage System (HPSS) with the Distributed File System (DFS) represents a state of the art in file systems technology. Another example is Coda, a file system which takes the notion of file system caching one step further allowing for network attached devices to be detached, while the work on a data object "hoarded" in the device's cache may continue unabated. Objectivity/DB, Oracle and UDB represent similar examples of state of the art in databases, with Java interfaces provided for coupling to the emerging world of pervasive computing. Widely adopted solutions that arise from the domain of personal computers often lack the elegance and comprehensive nature of, say, DFS, but they are adequate for small business, departmental, and domestic use. Solutions developed in the UNIX world are often adopted by institutions that have to manage a very large number of PCs. But with the exception of AFS none of these solutions has ever been scaled to a world wide file system, and AFS itself is restricted to about 160 cells and at most about one million principals. Database services are usually restricted to a single enterprise; the World Wide Web and its several search engines is a notable exception. These perhaps represent a state of the art in terms of their ability to handle a very large number of simultaneous requests.
Grand Challenges. The major challenge is to evolve and scale technologies such as file systems, massive data storage systems, databases, transaction engines, and, last but not least, hardware technologies that underlie the persistent storage framework of pervasive computing so that the data would become pervasive too. In other words, to assure that network principals can access data from every point and at all times within the network, assuming that they would present appropriate credentials to data servers and that they would be authorized to access the requested data. The data streams themselves would then have to be encrypted and signed so that only the authorized principal could decrypt the data stream and be assured of the integrity of the transaction and the data itself. It is clear that data management issues cannot be separated from privacy and security issues. The reason we need privacy and security in the pervasive computing context is because we need to protect data.
The universal data space of pervasive computing must be navigable. This implies among other things a consistent naming strategy: names that designate locations of all data objects must be referentially transparent. But data objects need to be richly annotated too, so that they can be located with ease. It is clear that file system and database technologies will need to merge at some stage. The pervasive computing data space must be scalable too. It should be possible for hundreds of millions, even billions, of network principals of various sizes to access data, sometimes even the same data object. This can be accomplished only if data are replicated and cached in strategic network locations. But first and foremost, mechanisms and tools that are currently used to access distributed file systems and distributed databases need to become more lightweight and more portable. For example, instead of implanting a Coda (or similar) client into a UNIX or NT kernel, we need to implant it into the Java Virtual Machine, and thus make access to such a distributed file system universal, portable, and lightweight.
Economic Impact. The pervasive data space is a fundamental component of pervasive computing. Its development and deployment will enable new types of economic activity such as information trade and commercial information storage. Already today we can observe the beginnings of this new economy. Digital music on the World Wide Web is a prime example; video on demand is another. The common feature of these and similar activities is that data (a movie or sound recording) are exchanged among network principals for money. The next step will be a provision of commercial storage space to network principals. Even today many institutions out-source their data management to external agents. For example the Department of Health of the Government of Queensland in Australia stores all its data at IBM headquarters in Brisbane. The data are accessed over the Internet and processed on the Department's local devices. In effect IBM has become a commercial Data Bank for the Government of Queensland. Given appropriate resources and a suitable technology they could provide data banking services to many other principals within the state and beyond. There is really very little difference between safekeeping money and safekeeping data. Today money is data and data is money. Tomorrow, as the economy becomes entirely cashless, there will be only data.
5. The Indiana Pervasive Computing Research Initiative | Table of contents | 7. IPCRES and Economic Development in Indiana
Copyright© 1999, The Trustees of Indiana University