Frontiers of Engineering: Reports on Leading-Edge Engineering from the 2001 NAE Symposium on Frontiers of Engineering (2002)

S. MUTHUKRISHNAN

AT&T Laboratories, Florham Park, New Jersey

Rutgers University, New Brunswick, New Jersey

Wireless communication has been around for a long time (e.g., commercial radio, satellite broadcasting, two-way radio systems, etc.). The first commercial wireless telecommunications networks were analog cellular systems; in the past decade, most of these systems have become digital (so called 2G) systems, and cellular traffic has increased dramatically. We now have four major cellular companies (cellcos) (Verizon Wireless with 27.9 million customers, Cingular with 20 million, AT&T Wireless with 18.8 million, and Sprint PCS with 11.2 million) and a dozen other large companies. There are an estimated 100 million wireless subscribers in this country, which is still fewer than the number of wire-line customers. Cellcos offer telephone service, voice mail, one- or two-way short messaging service (SMS), enhanced directory services (e.g., concierge service by AT&T Wireless), and E911 (rough locating of emergency callers). In addition, most cellcos offer basic data service (e.g., PocketNet by AT&T Wireless and Wireless Web by Sprint PCS), which give cell-phone users access to a

limited amount of information about movies, stock quotes, news, and sports. There is a data network—Cellular Digital Packet Data (CDPD)—that powers wireless modems for PC, laptop, and personal digital assistant (PDA), such as Palm or PocketPC, users and supplies data at roughly 10kbps (9.6, 13.3, etc.). There is also a fledgling convergence of PDA and phone systems—Kyocera produces PDAs that are also cell phones, and Handspring is a PDA with a module extension that makes it a cell phone. Overall, estimates of data users do not exceed 5 million nationwide. Internationally, technology and data rates are much the same as in the United States, but usage capacity and usage patterns differ. Wireless customers outnumber wire-line customers in a few European countries, and SMS is widely popular. Developing countries are adopting cell phones aggressively. China has more than 80 million subscribers; almost 40 percent of people in the Czech Republic subscribe to cell phones; and DoCoMo in Japan has 25 million subscribers to its data/information service called i-mode, which was launched in February 1999 (DoCoMo, 2001).

Most people believe that the cell phone market has just begun to take hold and that it will grow significantly; the enormously successful Internet is expected to be a new market for wireless technology. The world of the future is envisioned as a world in which users will be able to communicate anywhere and everywhere by voice (e.g., telephone calls) and data (e.g., text, audio, images, etc). Governments are providing incentives to encourage the growth of wireless communications to realize this vision, which calls for increased high-speed capacity. The governments of various countries in Asia, Europe, and North America have provided frequency spectrum for commercial development and auctioned off frequency licenses. Major cellcos are in various stages of building next-generation wireless networks (3Gs, with a possible evolution path via 2.5G) using this spectrum. Some debate has arisen about the advantages of different physical-layer technologies (e.g., TDMA, GSM, GPRS/EDGE, variants of CDMA, etc.). Vendors are in the process of producing the network elements (i.e., base stations, switches, etc.) to support the technologies service providers will need, and many technology trials have been conducted. Japan’s DoCoMo is doing a market trial of 3G systems, and AT&T Wireless is doing a market trial of 2.5G systems. Many cellcos in Asia (e.g., Japan, Korea, and Taiwan), Europe, and North America have announced plans for market trials and the deployment of 3G (2.5G+) systems over the next five years. These systems promise capacities of a few tens of kbps (50+ for GPRS) to a few hundred kbps (350+ for 3G) per cell, which will be significantly larger than the capacity of existing 2G systems. In short, we are on the threshold of building a new telecommunications infrastructure for large-scale use.

Major communication and services infrastructures (e.g., roads, bridges, railroads, telephone networks, radio broadcasting networks, cable networks, the Internet, etc.) are built only once every few decades. It is helpful to keep this perspective in mind when thinking about the next-generation wireless network.

Cost is an important consideration. Cellcos worldwide have spent hundreds of billions of dollars up front to acquire licenses to use spectrum, a surprisingly expensive “access cost” for developing a service in the historical context. Hence, market share will be a crucial guide to the nature and development of services. Technologically, wireless access (transport as well as backbone networks) will handle both real-time voice (telephone calls) as well as data (real-time streaming or nonreal-time) traffic. This will require more sophisticated networking than the existing Internet and, consequently, will pose new challenges. Emerging wireless networks will have to handle a large number of devices, and, because customers are likely to own more than one wireless device (e.g., cell phones, PDAs, laptops, and other appliances with wireless modems, etc.), 3G networks will have to handle many more customers than existing networks. Thus, scalability will be a central concern. One result will be that next-generation wireless networks may need larger address space than the space currently available in the Internet and, hence, may deploy IPv6¹ ; this will present another set of challenges because IPv6 is an emerging technology. Furthermore, end devices are likely to be diverse. Phones, computing devices, media, and gaming devices networked by wireless means will have different form factors, displays, and computing power and will understand varied protocols and data formats. Providing suitable service across heterogeneous wireless devices presents another challenge. In addition, the wireless channel is unpredictable, with variations that depend on the mobility of the user and the terrain, which affects perceived bandwidth and may lead to intermittent connectivity. Special care will be necessary in handling the application as well as the user to ensure that service is adequate with any application. Finally, security and privacy concerns will be critical, perhaps even more urgent than in existing Internet and telecommunications infrastructure because the air interface is more accessible to snoopers than other access methods, and users will have a large “footprint” on the network because of their mobility.

To summarize, emerging wireless networks will involve the deployment of cutting-(bleeding?)-edge technologies that will be converged and complex, a marvelous engineering achievement. In general, the deployment of new networks will provide opportunities for initiating new value-added services. In the current context, massive investments that have been (and will be) made will provide additional incentives for companies to develop viable revenue sources, and hence, to focus on services and applications.

I will approach the subject of applications and services for wireless users in emerging networks primarily from the point of view of large, cellular

telecommunications companies (cellcos), which will play a major role in the development of the next-generation wireless infrastructure and have already made a sizeable investment. Realistically, however, no one cellco will become the one-stop portal for all services of interest to wireless users. It is reasonable to expect that third-party companies will provide niche services. In the future, the industry will have to resolve the tension between services managed by cellcos (within the so called “walled garden”) and services managed by third-party companies and find revenue-sharing mechanisms for third-party services. The following description of services and applications in emerging wireless networks will address the technological, as well as the historical context. I begin with detailed descriptions of two services—enhanced messaging and location-aware services—as a way of identifying some of the issues a services architecture will have to address. Then, I will provide an overview of other considerations for cellcos designing a services architecture.

Current wireless systems provide simple, nonreal-time messaging services, such as voice mail and one- or two-way text messaging using SMS and paging; instant messaging is a popular real-time application. Emerging wireless telecommunication systems envisage messaging applications with much richer media, involving audio, video, Web, and text messages. Here is a sample scenario for multimedia messaging.

Alice gets off a plane and turns on her phone (the use of 3G systems will continue to be prohibited on airplanes!). She has an email from Bob containing the system drawing in image format and a note that all changes have been made as previously agreed upon. She trusts Bob, so without checking the image, she forwards it to the product group with a voice memo. As she waits for her baggage to arrive (3G networks will not reduce baggage delivery time!), she checks her personal email and sees a message from her supermodel sister, Carole, containing a cute animation video clip of “Dilbert.” She watches the clip and forwards it to her colleagues and friends. She also responds to her sister with an archived audio clip of the “Top Ten Reasons Why Models are Clueless” from David Letterman. Meanwhile, she receives an SMS message notifying her of a birthday e-card from Erin that was sent to her home PC earlier; she connects to the associated Web site, listens to part of the audio, which happens to be her favorite song, and skips to the end. The e-card image cannot be rendered on her phone legibly, so she forwards it to her home PC. As she walks to the exit, she receives an instant message from the limo driver waiting for her. She responds and is directed to the limo. Finally, she sits back massaging her aching eyes, wrists, and shoulders—airplane rides and 3G handsets are still hard on your joints!

This simple, fictional scenario captures some of the expectations of future wireless systems. First, to engage users, applications will have to be highly relevant, simple to use, and fun. Current mailers would be challenged to meet

these requirements. The first problem is to avoid redundant data transfers. For example, Alice forwards Bob’s image attachment without viewing it. Therefore, there is no reason for the image attachment to be downloaded from the mail server to the handset and resent from the handset to the mail server for forwarding. This would waste uplink and downlink bandwidth. The second problem is data transcoding. Wireless handsets are likely to remain heterogeneous, with different abilities to decode data in different formats or to render them meaningfully at different granularity. Current mailer systems do not provide a way for clients to negotiate from one level of detail to another or to request conversion from one format to another. This is likely to be a major problem. Either fairly uniform handsets will have to be introduced, or the content will have to be customizable for any of the available devices, neither of which is a clean solution to the problem. Negotiating the desired data format is a network-level challenge; transcoding from one format to another is a data-management challenge.

Third, the problem at the back end is duplicate data, or the “curse of forwarding”; data is duplicated many times over because personal users tend to forward their favorite attachments to friends, family, and colleagues. In commercial settings, a manager may send a notice to several employees, or a portal may send an advertisement clip to several customers. In current mailer systems, the forwarded documents are predominantly text documents. If we extrapolate the phenomenon to bulky multimedia content (e.g., audio and video clips) over the wireless networks, where multimedia content is expected to be extensive, and mail management is performed by cellcos with tens of millions of users, the problem takes on a different dimension. Finally, there is the problem of “atomic content.” Current mailer systems do not allow the mobile user to see a “synopsis” of the multimedia attachment (e.g., a summary of the content, rather than a display of content type) or to perform light editing (e.g., searching to an interesting portion of the text and splicing in a figure or a Web link) without downloading the entire attachment. Thus, the multimedia content is essentially “atomic,” that is, it allows for no flexible, client-server manipulation. Unless this changes, this will result in a very heavy handset (e.g., all software and resources for multimedia manipulation must be on the handset) and could possibly lead to a poor overall user experience.

The problems I have described above are mainly data management problems, and they are not difficult to solve, in principle, but they will require clean engineering solutions. For details on how to store duplicate, multimedia dataspace efficiently, how to perform on-demand content-type negotiation for transcoding, how to enable message sending by proxy to avoid redundant data transfers, and streaming (versus downloading) of multimedia data in the messaging context, see Nelson and Muthukrishnan (2001).

Now let us focus on a specific problem inherent in the scenario described above, namely, maintaining the presence of a user on the network. Current cell phone systems use home location registers (HLRs) and visiting location registers

(VLRs) to maintain device status information.² For every cell phone, the HLR stores account status, presence information, and the mobile switching center (MSC) currently serving the device. A VLR, a component of an MSC, stores status information about devices in the MSC’s service area. For example, a VLR often stores an estimate of the device’s location to reduce the paging required to complete an incoming phone call. Instant messaging uses subscriber presence information to update buddy lists and determine when instant messages can be delivered. For conventional Internet use, determining presence includes determining whether the user is logged in and whether the user is actively using the computer (e.g., see <http://www.jabber.com/pdf/Jabber_Server_White_Paper.pdf>). A weakness of current technology for maintaining and serving users’ status information is the focus on device status rather than on customer status. Many customers own more than one mobile wireless device (e.g., a cell phone, an advanced pager [such as a RIM Blackberry], and a PDA with wireless connectivity [such as a Palm VII]). Some customers own more than one cell phone (e.g., a work cell phone and a personal cell phone). Hence, wireless application infrastructure must store and serve information about the user, rather than the device.

Another specific problem, associated with the multiple presence of a user, is maintaining user preferences for notifications. Customers may want to use their devices in conjunction with other communications mechanisms. For example, users might desire SMS notifications when email is delivered. If the user is near one of his or her wire-line phones, he or she might prefer to receive voice messages on the wireline device. Similarly, if the user is active on an Internet connected computer, he or she might prefer to receive messages on the computer. Also, a user named Joe on hotmail who receives an instant message request might be called John on a 3G phone and would prefer to receive an SMS message as an alert. To deliver these services, the wireless applications infrastructure will have to provide a mechanism for users to set their preferences for handling messaging events as logical rules, which will have to be maintained and applied by the service provider.

Mobility is a unique feature of wireless networks. Therefore, a user’s location at any given time is a unique resource. Offering user services based on current, posited, or desired location is a tantalizing idea. I will address three issues: location-aware services; determining a user’s location; and an architecture for providing location-aware services.

The simplest location-aware service would be providing geopersonal information, that is, providing information about a user’s interests based on location.

The information may be surreal (e.g., “Where am I?”) or more practical (e.g., “Where is my child, spouse, or FedEx delivery agent now?”). Other typical information would involve locating closest businesses (e.g., hardware stores), entertainment (e.g., movie theaters or restaurants), traffic (e.g., road conditions), and so on.

The second issue is location determination. One approach is to coopt the user to enter a certain location, which would be inconvenient; therefore, I will focus on automatic localization without user’s aid. The Global Positioning System (GPS) provides a satellite-based solution that is accurate but does not work indoors or in dense urban areas. Furthermore, GPS must be incorporated into the handset, an expense cellcos may prefer to avoid. Some progress has been made toward creating low-cost chip sets and assisted or differential GPS solutions to overcome current coverage problems. There are also network-centric location determination methods by which a user can query network elements and use radio-layer parameters. These methods include: basic cell-ID that identifies the cell region of the user; and refined methods, including AOA (angle of arrival), E-OTD (enhanced observed time difference), signal attenuation, and TDOA (time difference of arrival). These methods all require that network elements be accessible, be “queryable,” or be able to dump traffic logs frequently; they also require vendor support and would tax already overburdened network elements. In general, these methods are preferable only when handsets are “dumb,” that is, not equipped with a powerful processor and a sophisticated operating system. For smart handsets, a scalable solution is to rely on the back channel cell-ID notification and translate cell-ID to physical location with the help of the cellco. This works very well and suffices for most geopersonal information services (Muthukrishnan et al., 2001). In a related note, E911 is a government-enforced program that localizes emergency callers; whatever infrastructure cellcos put in place must meet this demand, but only for in-progress cell phone calls and only for a small number of calls (statistics put the number of 911 calls per year nationwide at 50 million, while the number of cell phone calls per day is likely to be in the hundreds of millions [CTIA, 2001]). Hence, the infrastructure need not provide a scalable solution for location-aware services for significantly more data users particularly for location tracking. See Hightower and Borriello (2001) for a recent comparison of location-determination technologies.

A third issue is the development of an architecture for providing location-aware services. Although I will focus on data users with smart handsets, the discussion also applies to voice users and data users with other handsets, with suitable restrictions. The first approach is a location-server architecture. In its simplest form, the user’s location information is stored in a database together with customer information. This approach is dynamically populated by the location-determination technologies. The mobile user has the ability to direct any information service provider (e.g., Yahoo, Mapquest, etc.) to access the location information from this server. A number of details will have to be

resolved, including how requests can be authenticated at the location server, how adequate privacy and security can be provided, and appropriate interfaces for location information suited to different applications. The second approach is a proxy-based architecture. In its simplest form, this consists of a proxy with which mobile clients communicate. Mobile clients pass their “location” information to the proxy, which not only translates it into geographic location, but also acts as a conduit to third-party information service providers by converting user requests to geocoded requests (e.g., by rewriting the URLs). Again, a number of issues will have to be addressed: how security, such as privacy and authentication, can be provided; where the proxy should be located; and appropriate interfaces for location information suited to different applications and third-party information service providers. Other architectures are possible, including local proxies.

There are also more advanced location-aware services. Instant messaging has recently expanded into wireless communications. In this context, standards organizations have recognized that location may be used for other purposes, such as sending messages to someone only if they are less than a mile from the user, etc. (see <http://www.instantmessaging.org/>). This would require tracking mobile users. Some location-aware services could include changing the ringing tone/volume, based on where the user is or letting the user know when he/she is at an exit on the highway. Two enticing applications are geographic push, that is, notifying or messaging any user in a geographic area, and geographic bookmarking, wherein users “bookmark” locations, manage them, and navigate using them. Location tracking would tax the scalability of the infrastructure for location-aware services, but geographic push may be considered a basic component. Geographic bookmarks would entail careful data management systems in the back end. Finally, portals might be personalized to the user’s location.

Besides enhanced messaging and location-aware services, next-generation wireless systems would enable a suite of services: administrative services (e.g., profile, preferences, purchases, personalization, security), content services (e.g., providing a Web space for downloads, stored content, coupons), personal information management (e.g., schedules, contacts, synchronization with remote devices), and transaction services (e.g., prepayments, credit, cash withdrawals). Users’ expectations and experiences with wireless applications will evolve over time (Henry and Ziang, 2000). In addition to the basic services described above, cellcos are likely to support “walled garden” services and applications and provide a branded portal with associated services, such as searching. Finally, cellcos are likely to provide mechanisms for corporate Intranets, including VPNs, and for external net and third-party services. Some of these issues are discussed in

3GPP (2001). The schematics of the services architecture and a description of the flow of controls and data for various services are provided below.

The entire services architecture will have to be deployed at the scale described above. Therefore, it will have to be distributed and duplicated with careful load balancing. Database management will be critical, and the entire system will have to be engineered and maintained: including provisioning, billing, hardware/software/configuration management, fault and performance management, and data centers and customer care. All of these functions are likely to be more challenging than they are for existing communication networks. The entire endeavor will have to be of “telco quality,” that is, with tiny fault probability, which will necessitate failback options. Finally, the services architecture will have to be built for the short-term, as well as for long-term evolution. Roaming among carriers and (inter)national markets, as well as interoperability of systems, services, and interfaces, should be considered for the short term. Telematics (i.e., wireless access to car users) will be of short-term to midterm interest. In the long term, wide area telecommunication infrastructures will have to interface with local, home, and personal area networks to form unified networks. Finally, service providers are likely to develop novel applications, much as they have on the Web. Whatever services architecture is put into place, it will have to be capable of evolving, as necessary.

Much of the discussion thus far in the media and in industry has been about physical-layer access technologies, spectrum allotments, and “killer apps” in next-generation wireless systems. I have presented some issues related to the design and deployment of a (conservative) services infrastructure. The challenges are likely to be of the “computer science and engineering” variety, namely, database management, scheduling and load balancing in distributed systems, user interface design, security, privacy, and authentication, on a grand scale.

If we look beyond engineering, a new communication and services infrastructure is being put into place, which means we have an opportunity to specify public needs that must be addressed. For example, public safety, crisis and emergency management, services for the physically challenged, and so on can be championed and implemented. From a scientific point of view, a large number of mobile users with a variety of handsets equipped with high-quality wireless connectivity represent a unique capability of gathering vast quantities of data in a distributed, real-time manner that can be integrated with other data sources.

The best way to take advantage of this previously unimaginable opportunity has yet to be determined. Market and industry may have to evolve to isolate certain “free content” to provide third-party mobile application developers an opportunity to add services atop the basic infrastructure. For example, local governments could deploy traffic and weather monitoring nodes and network

them by wired and/or wireless means, and then make the data feed available to all service providers. The new infrastructure will make it easier for small third parties to provide geography-specific applications.

New communication infrastructure is also a vehicle for innovative social use. For example, artists might plant “graffiti” in the wireless world that can only be seen by users on their devices based on geographic proximity, or a large-scale symphony might be orchestrated by appropriately dialing cell phones in the audience and using their ring tones. The possibilities are limited only by the imagination.

Finally, the question arises whether devices and users can be addressed based solely on their geographic attributes, such as location and direction of movement. Answering this question will require a higher level of abstraction akin to content-addressable networking on the Web. It remains to be seen what kind of applications this will lead to.

Thanks to Rittwik Jana, Ted Johnson, and Andrea Vitaletti, my colleagues working on location-aware services; Darin Nelson and Daryl Bunce at AT&T Wireless for multimedia messaging system design; Nathan Hamilton for sharing his expertise; and Robin Chen, Paul Henry, and Rob Calderbank for their discussions and perspectives.

CTIA (Cellular Telecommunications and Internet Association). 2001. Available online at <http://www.wow-com.com/news/publications>.

DoCoMo. 2001. Subscribers for i-mode Service of NTT DoCoMo. 2001. Available online at <http://www.nttdocomo.com/i/i_m_scr.html>.

Henry, P., and Z. Ziang. 2000. A Subjective Survey of User Experience for Data Applications in Future Cellular Wireless Networks. AT&T Technical Memorandum.

Hightower, J., and G. Borriello. 2001. Location systems for ubiquitous computing. IEEE Computer (Special Issue), August 2001:57–66.

Muthukrishnan, S., R. Jana, T. Johnson, and A. Vitaletti. 2001. Location based services in a wireless WAN using cellular digital packet data (CDPD). Pp. 74–83 in Proceedings of the ACM International Workshop on Data Engineering for Wireless and Mobile Access (Mobide). New York: Association for Computing Machinery.

Nelson D., and S. Muthukrishnan. 2001. Design issues in multimedia messaging for next generation wireless systems. Pp. 98–103 in Proceedings of the ACM International Workshop on Data Engineering for Wireless and Mobile Access (Mobide). New York: Association for Computing Machinery.

3GPP (Third Generation Partnership Project). 2001. Available online at <http://www.3gpp.org>.