Copyright 1999 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

This article was published in the June 1999 issue of
IEEE Personal Communications.

ABSTRACT

The integration of General Packet Radio Service with IS-136 networks will provide IS-136 users and operators with many benefits. In addition to the existing voice services on circuit networks, the mobile user with an appropriate terminal now has direct access to services on packet data networks through GPRS. Of particular importance are services based on the Internet Protocol (IP). Furthermore, compared with fixed IP networks, there is an added dimension stemming from user mobility, which results in mobility-specific requirements and mobility-specific services. In addition, GPRS protocols open up the possibility of global data roaming with Global System for Mobile Communications networks. From the operator perspective, the integration enables a controlled and flexible architecture transition to a network where data applications play an increasingly important role. The operator also benefits from the GSM economies of scale and from the evolution roadmap of GSM to the third generation. Finally, the above service advantages to the user translate into a competitive advantage for the operator. This article describes the architecture resulting from integrating IS-136 with GPRS and discusses the service evolution as well as network evolution that can be envisioned from such integration.

GPRS and IS-136 Integration for Flexible Network and Services Evolution

Stefano Faccin, Liangchi Hsu, Rajeev Koodli, Khiem Le, and Rene Purnadi
Nokia Research Center

With the advent of the World Wide Web, the Internet has grown beyond reasonable imagination from a network that was intended for collaboration among a selective group of researchers to a network that is rapidly influencing our lives by changing the existing paradigms of communication and opening new avenues. Specifically, the Internet has paved the way for data networking, in which networks based on the IP packet-switched model will support voice, data, and video within a unified network infrastructure, bringing about economies of scale to end users as well as to operators. Meanwhile, the cellular market continues to grow at an impressive pace. Not surprisingly, the volume of cellular data devices is expected to grow at a phenomenal rate as well, as evidenced in Fig. 1. Thus, it appears that the cellular data sector, which is expected to benefit from growth in both the Web and cellular areas, promises to be an exciting area for technology innovation and market potential.
Existing cellular systems (up to second-generation systems) provide to varying degrees access and connectivity to packet data networks. However, since cellular systems were originally designed for conversational voice applications, the connection mechanism is based on the circuit-switched paradigm; that is, a dedicated circuit is reserved on the air interface, although the traffic has a bursty nature, which results in inefficient radio resource usage. In addition, the connection model is based on dial up modems, which results in high set up latency time. Another type of data-related service is short message services (SMS). SMS use radio resources more efficiently, but have message length and overall throughput limitations.
To address these inefficiencies, two packet radio schemes have been introduced in cellular so far: cellular digital packet data (CDPD) and GPRS. Both are based on a packet paradigm. CDPD can be used with Advanced Mobile Phone System (AMPS), IS-95 and IS-136, while GPRS was originally defined for GSM. In the IS-136 evolution to the third generation, a major thrust is the provision of enhanced packet data capabilities. The two components of the capabilities are a higher bit rate over the physical layer, and upper layers better tailored for packet data. Higher air interface bit rates are achieved through either a modified 30 kHz modulation scheme (IS-136+ which achieves a peak rate of ~60 kb/s) or a new 200 kHz modulation scheme in Enhanced Data rates for GSM Evolution (EDGE), which achieves a peak rate of 384 kb/sec). Until the recent past, an evolved version of CDPD was considered for the upper layers. More recently, a decision was made to adopt a GPRS-based solution rather than a CDPD one. This article describes the fundamental features of an integrated IS-136/GPRS system as well as the services enabled by such a system, and the possible future evolution from the network and service perspectives. The integrated IS-136/GPRS system will be referred to as GPRS-136.
The rest of this article is organized as follows. The "Background" section provides background information on CDPD and GPRS, along with a comparison. The next two sections describe the salient features of the GPRS-136 architecture, as well as the services enabled by the architecture. The following sections consider the evolution of the network as well as services beyond GPRS-136.

Background

The CDPD System

CDPD was initially designed as an overlay system on top of the AMPS networks. Subsequently, it was adapted to IS-95 and IS-136 networks. Services provided are access to networks based on IP and Connectionless Network Protocol (CLNP).
Figure 2 shows a network view of the CDPD network [1]. The network nodes of CDPD are home and serving mobile data intermediate systems (MD-ISs) and the mobile database station (MDBS). Basically, intermediate systems are IP-capable routers that form the backbone of the CDPD network. They are responsible for relaying user data, network administration, and mobility information. The home MD-IS stores the mobile station profile, authenticates the mobile station, and provides the point of entry for IP datagrams destined for a mobile station, which are encapsulated and routed to the proper serving MD-IS.
There are two layers of mobility management in the CDPD network. The home MD-IS performs macro mobility management of tracking which serving MD-IS is currently serving the mobile station, while the serving MD-IS is in charge of the micro mobility management of tracking the mobile station down to the cell level.
The MDBS forms the physical interface between the baseband and radio frequency (RF) environments, and acts as a CDPD radio channel controller managing the radio resources. Over the air, the mobile end system (M-ES) accesses forward and reserve channels using digital sense multiple access with collision detection (DSMA/CD).
CDPD has its own set of databases (independent of the AMPS, IS-95, or IS-136 networks) for mobility management and subscriber profile information. The corresponding signaling protocol to manage the data structures is also specific to CDPD and unrelated to IS-41, the counterpart signaling protocol in the underlying AMPS, IS-95, or IS-136 cellular network. Thus, a CDPD-capable cellular terminal is expected to perform two independent mobility management processes, one for CDPD and one for the cellular system.

The GPRS System

GPRS is a new GSM service introduced in order to provide more efficient access to packet data networks from cellular networks. GPRS is based on packet transmission over the air interface and in the network, and therefore allows more efficient resource utilization. GPRS is particularly well suited to carrying Internet traffic, which is often bursty with fluctuating data rate requirements. GPRS defines a general framework for cellular connection to a variety of packet data networks. Currently, support for IP networks and X.25 networks is included in the specifications.
GPRS introduces a totally new backbone network based on IP, composed of new packet network nodes and traditional packet Internet nodes (router, domain name server or DNS, servers, firewall, etc.). Figure 3 provides a network view of regular GPRS, as designed for GSM. GPRS adds two new main network elements to the existing infrastructure: the serving GPRS support node (SGSN) and the gateway GPRS support node (GGSN). These elements interact with each other and with the existing cellular network elements over a set of new interfaces. In particular, two new interfaces are standardized: a Gb interface between the BS subsystem (BSS) and the SGSN, and the (optional) Gs interface between the SGSN and the mobile switching center (MSC).
SGSN takes care of terminal mobility and authentication functions, and is connected to the BSS over a frame relay network on one side and to the GGSN over an IP backbone network on the other. GGSN, in turn, provides connections and access to external networks. As regards the external IP network, GGSN can be seen as performing common IP router functions.
Other network elements added to the infrastructure are the border gateways (BGs), used to interconnect two intra-PLMN backbone networks via an inter-public land mobile network (PLMN) backbone network, and the firewalls (FWs), used to protect the PLMN from unauthorized accesses.
SMS is supported by a direct logical connection between the SMS centers and the SGSN (Gd interface).
In the remainder of this section we describe in more detail the following relevant features of GPRS:

GPRS mobility management
GPRS Tunneling Protocol (GTP) and Packet Data Protocol (PDP) context
Quality of service (QoS) support

GPRS Mobility Management -- GPRS has its own mobility management, separate from GSM mobility management. However, unlike in CDPD where mobility management is independent of IS-41, the GPRS mobility management protocol is derived from its circuit counterpart, GSM Mobile Application Protocol (MAP). Databases for mobility management and subscriber information reside in the SGSN (which has a visited location register or VLR-like functionality) and home location register (HLR). International Mobile Station Identity (IMSI) is used as the primary key in accessing the network databases, and the GPRS-related subscriber information in the HLR can be seen as an extension of the GSM information. Also, the GPRS security scheme is derived from GSM MAP.
GPRS mobility management consists of two levels: macromobility to track the current serving SGSN, and micromobility to track the terminal location down to the routing area or cell level. Macromobility management involves profile downloading from the HLR to the serving SGSN and GTP tunneling/PDP context management, explained in the next section. At the micromobility level, the mobile station (MS) is tracked depending on its mobility management state. The IDLE state represents a user who cannot be reached by the GPRS network. A GPRS Attach procedure allows the MS to be in READY state in which its location is known at the cell level. The MS slips to STANDBY state when there is no active transmission. In this state the MS is tracked down to a routing area, which is a subset of a GSM location area. In order to enable the transfer of data packets, the MS must activate a PDP context, explained later in the following section.
A GPRS MS can operate in one of three modes of operation:

Class A: The MS can attach to both GPRS and other GSM services, and the MS supports simultaneous operation of GPRS and other GSM services.
Class B: The MS can attach to both GPRS and other GSM services, but the MS can operate only one set of services at a time, and is capable of monitoring one paging channel at a time.
Class C: The MS can exclusively attach to GPRS services.

The mobility management coordination issue is how to page for GPRS and GSM services when the terminal only listens to one paging channel, as in the class B mode of operation. In GPRS, a GPRS Attached MS operating in class B will monitor paging by listening to the GPRS paging channel or GPRS traffic channel, if it has been assigned one. This behavior will be referred to as camping on the packet side.
GPRS has a network mode of operation (mode I) whereby mobility management is coordinated between GPRS and GSM through the Gs interface so that class B MSs can still be paged for voice or circuit-switched (CS) calls when camping on the packet side. In that mode, all MSC-originated paging of GPRS-attached MSs shall go via the SGSN, thus allowing network coordination of paging by the SGSN based on the IMSI. The SGSN converts the MSC paging message into an SGSN paging message. In addition, paging via the SGSN will be more efficient, because the SGSN will page in either a cell or routing area. A routing area is always fully contained in a GSM location area.
Another feature that saves radio resources is the combined CS and packet-switched (PS) mobility management procedures. For example, a combined routing area (RA)/location area (LA) update procedure takes place when the MS enters a new LA while the MS is IMSI- and GPRS-attached. The MS sends a single RA update request, indicating that a LA update shall also be performed. The SGSN interprets the request, executes the RA update, and forwards the LA update to the MSC/VLR.
Complete mobility management, including the location management procedures, is outlined in [2].
GTP Tunneling and PDP Context -- A PDP context is a two-way tunnel between the SGSN and the GGSN used to carry encapsulated user packets. When the attached user wants to send or receive data, he or she has to activate a PDP context. In IP, activating a PDP context means activating an IP address which will be used by the user as his/her IP address for that specific service. A GPRS attached user then can be associated with either a static IP address, assigned permanently by the home PLMN (HPLMN) operator at subscription time, or a dynamic IP address, allocated temporarily by the HPLMN or the visited PLMN (VPLMN) operator at the PDP context activation time by the GGSN.
Figure 4 shows the GPRS transmission plane. Between the SGSN and the MS, PDP data are transferred with the Subnetwork Dependence Convergence Protocol (SNDCP,) whereas between the SGSN and the GGSN, PDP data are routed and transferred with either the Transmission Control Protocol (TCP)/IP (for X.25 services) or User Datagram Protocol(UDP)/IP protocols (for IP services). PDP context activation leads to the establishment of an SNDCP tunnel and a GTP tunnel used over the radio between the SGSN and the MS and in the network between the GGSN and the SGSN, respectively. The user may have up to 15 simultaneous PDP contexts active at a given time. These contexts can even serve different PDP protocols (IP, X.25, etc.).
QoS in GPRS -- QoS requirement variations depending on the type of the connection (Internet, Internet service provider or ISP, corporate, etc.) and the purpose of the connection (e-mail, multimedia, Web access to the Internet, browsing corporate databases, etc.) require that GPRS provide a flexible and general QoS mechanisms. GPRS specifies a QoS profile for each PDP context, defined by various attributes such as the delay class, indicating the scheduling order of data packets belonging to different subscribers and PDP contexts, and the mean throughput class, specifying the average rate at which data is expected to be transferred during the remaining lifetime of an activated PDP context. Other QoS attributes are precedence class (indicates drop precedence), reliability class (indicates error rate tolerance), and peak throughput class (indicates maximum data transmission rate).
The SGSN maps the QoS profile defined for a PDP context into the appropriate radio link control (RLC)/medium access control (MAC) radio priority level and indicates to the MS what priority value it should use in uplink access. The MS uses this value and the reason for uplink access (i.e., signaling or data transmission) in the uplink access request. The BSS then determines the radio access precedence based on the information supplied by the MS.

A Comparison between CDPD and GPRS

At a high level, CDPD and GPRS are both based on the same fundamental principles: tunneling of packets to the destination, and two levels of mobility tracking, macro- and micromobility. However, GPRS has some advantages over CDPD: QoS support, coordination between circuit and packet data mobility management, and dynamic IP address allocation.
QoS support is a crucial requirement for services beyond the traditional delay-tolerant packet data services. GPRS has taken QoS into account and allows specifying QoS profiles with the following attributes: service precedence, delay, reliability, mean throughput, and peak throughput. GPRS also can provide multiple PDP contexts, each with its own QoS profile, for various packet data services.
In GPRS, coordination between circuit and packet mobility management results in a more efficient paging mechanism; that is, the mobile station needs to listen to only one paging channel, and the paging area is minimized.
Dynamic IP addressing is essential to support the large number of cellular users, without necessarily requiring the use of IP version 6 (IPv6). It allows faster deployment of packet-based services to cellular networks using the existing IP version 4 (IPv4).
In addition, GSM has gained worldwide acceptance and deployment, and consequently GPRS will benefit from the GSM economies of scale. GPRS can also provide IS-136 users with global data roaming capability with GSM/GPRS. GPRS is part of the overall GSM roadmap to the third generation, and therefore is expected to be enhanced with additional capabilities. In particular, GPRS will work with the future EDGE modulation and its higher bit rate.

GPRS-136 Architecture

General Description

The primary goals of GPRS-136 architecture are to:

Allow for the adaptation of GPRS protocols and systems into the IS-136/IS-41 environment in a manner that minimizes changes to both systems
Leverage work already done in the Telecommunications Industry Association (TIA) on MAC and physical layers
Leverage development work on GPRS systems within the GSM community

Consequently, the architecture consists of an IS-136+ physical layer, and upper layers as in GPRS, except for the MAC layer, which is derived from existing TIA work. To the GPRS layers, this MAC layer should look like the regular GPRS RLC/MAC layer (i.e., provide the same services).
An issue specific to GPRS-136 is that the IS-136 network is based on the IS-41 mobility management protocol, rather than GSM MAP. In order to minimize changes to both systems, it was decided to have a separate HLR (at least functionally) for GPRS. Thus, there are two parallel systems, each with their own mobility management protocols. However, mobility management coordination is kept through the Gs interface and tunneling (Fig. 5).

Mobility Management in GPRS-136

The main function of the mobility management entities, as defined in TIA/EIA 136–336 [3], within a GPRS-136 mobile station is to support the mobility of user terminals, such as informing the network of its present location and providing user identity confidentiality. The GPRS-136 packet data network is obtained by combining IS-41 CS network elements with GPRS packet data network elements. The interaction between the MS and the IS-41 network elements when the MS is obtaining service from the packet data network elements is achieved by the technique of tunneling. This tunneling concept is a key element of the mobility management model for GPRS-136. A Class B 136 MS interacts with an IS-136 MSC by tunneling IS-136 signaling messages transparently through the SGSN. The SGSN does not interpret the IS-136 signaling messages flowing back and forth between the MS and the MSC. Combined procedures such as combined RA/LA update are no longer possible, since they require interpretation of LA update by the SGSN, which is contrary to the tunneling principle.
There are two entities, the GPRS mobility management (GMM) and IS-136 mobility management (136MM) entities, that are maintained as parallel state machines which are taken together to implement mobility management in GPRS-136:

The GMM entity performs procedures for packet data mobility management while obtaining service from the GPRS-136 packet system. The GMM entity in a GPRS-136 MS performs these procedures by interacting with the packet data network elements.
The 136MM entity performs procedures for mobility management, primarily for non-GPRS circuit services. The 136MM entity in a Class B 136 MS performs these procedures by interacting with the IS-41 network elements by tunneling IS-136 signaling messages transparently through the SGSN.

Interaction between the GMM and 136MM entities is necessary. As shown in Fig. 6, this interaction is typically in the form of a request from 136MM to GMM to initiate a GMM procedure (e.g., attach, RA update, detach, suspend) and the response from the GMM entity to such a request triggering a 136MM procedure or action. The GMM entity is expected to notify the 136MM entity when certain GMM events of interest to 136MM take place. Similarly, the 136MM entity is expected to notify the GMM entity when certain 136MM events of interest to GMM take place. Although GMM and 136MM are closely related entities in a GPRS-136 MS, when the MS is on a packet system the GMM entity is the primary manager of mobility with respect to the packet system. The 136MM entity is a client of the logical link control (LLC) layer and uses the services offered by it to tunnel IS-136 signaling messages.
Tunneling of Signaling Messages -- As mentioned before, in order to facilitate the use of GPRS network elements in non-GSM networks, SGSNs may support the tunneling of non-GSM signaling messages between a non-GSM MSC/VLR attached to the SGSN using a facility called tunneling of signaling messages (TOSM) [2]. This tunneling scheme is generically applicable to use GPRS in conjunction with other non-GSM cellular systems. TOSM is supported by using two LLC service access points (SAPs) -- high-priority signaling tunnel (HPST) and low-priority signaling tunnel (LPST) -- between the SGSN and the MS. The TOSM procedure is illustrated in Fig. 7. When the MS wants to communicate with the MSC/VLR, it sends the specific signaling message to be tunneled on one of the two LLC SAPs used for TOSM. The SGSN identifies the MSC/VLR to which to forward the message based on the routing area identity (RAI) where the MS is located, and forwards the message to the identified MSC/VLR along with the IMSI of the MS. On the other hand, if the MSC/VLR wants to communicate with the MS, it sends the specific signaling message to be tunneled to the SGSN associated with the MS along with the IMSI of the MS, and also indicates the priority of the message and whether LLC encryption shall be used or not. The SGSN sends the signaling message to the MS on the appropriate LLC SAP based on the priority of the signaling message indicated by the MSC/VLR. The SGSN provides just a transport service for the messages to be tunneled and shall not interpret the tunneling payload.

Services

Broadly speaking, there are two types of data services envisioned for wireless cellular networks: the services already prevalent in fixed IP networks, and those specific to wireless networks.

IP Fixed Services

In the fixed wireline IP domain (i.e., most of today's best-effort Internet), many "services" exist ranging from e-mail, ftp, and the Web to streaming multimedia applications including audio and video. Of these, the Web is the most dominant application in terms of both the amount of traffic it accounts for as well as the revenues it helps generate. It is worth noting that what is considered a commonplace application on most desktops and laptops is only slowly emerging in wireless devices. Also, a common feature of the existing services is that they do not require strong QoS support, although the streaming applications can greatly benefit from even a rudimentary form of QoS support. However, with the emergence of differentiated services [4], the Internet is fast evolving into a network which can support multiple classes of services with relative QoS support among the classes instead of absolute per-flow QoS. This development has a tremendous influence on service provisioning in both the wireline and wireless domains.

New IP Services

In addition to supporting services available in the wireline domain to individual users, many new services will be enabled by the untethered nature of cellular packet data services. Examples include warehouse inventory control, and automated metering for utility companies and toll collection on highways. Also, wireless virtual private networking will evolve in coming years.
There are other data services enabled by the mobility of users. For example, location-aware services are emerging as a major segment of the cellular market. In such services, the exact location of the user, which is determinable, for example, using the network-assisted Global Positioning System (GPS), is used to offer services such as traffic updates, directory services, emergency services, and weather forecasts, to name a few. For example, a mobile user may speak into his/her device "searching" for restaurant information, which may then be converted by a speech-text converter into pure data and transmitted to the network. The network in turn may relay the request to a directory server, which determines a list of restaurants using the location information supplied and provides the response back to the user. The response may be supplied back to the user in the form of either a textual display and/or voice. Such services can already be supported using cellular data, such as by the Wireless Access Protocol (WAP) architecture. However, with the introduction of GPRS, the data services can be built using the IP paradigm directly with or without an intermediary architecture such as WAP. More important, GPRS offers ubiquity in terms of coverage, so applications can run seamlessly regardless of the location of the mobile user. In addition, GPRS offers a higher bit rate, point-to-multipoint support, and efficient network usage for bursty packet-based applications.
Finally, cellular SMS, although not totally new, have lately become very popular, especiallyl in Europe. Such a trend will likely happen in the Americas and on other continents as well. GPRS will enable us to address the throughput limitations and allow further expanded SMS.

Future Evolution

There are multiple dimensions in the evolution. The first dimension is the physical layer of the radio interface. The second dimension, in principle orthogonal to the first one, involves the upper layers of the air interface and network. Closely associated with network evolution is service evolution.

Physical Layer Evolution

The next step after GPRS-136 is the adoption of the 200 kHz carrier EDGE BSs, which will be placed side by side with the current BSs. EDGE is a new modulation for 136 that will achieve even higher bit rates for packet data services (i.e., 384 kb/s). The adoption of the 200 kHz carrier BSs require more spectrum availability, although it can result in higher spectrum efficiency.
Future evolutions of IS-136 may contemplate a wideband time-division multiple access (W-TDMA) solution with 1.6 MHz carrier, providing data rates up to 2 Mb/s. This solution is also called indoor since the International Telecommunication Union (ITU) requirement for outdoor services is 144 kb/s with the indication that 384 kb/s is preferable, whereas providing a data rate of 2 Mb/s for outdoor services is not required. Another possible direction of evolution is the migration to wideband code-division multiple access(W-CDMA). Availability of spectrum is prerequisite to the deployment of W-TDMA and W-CDMA.

Upper Layers and Service Evolution

The new EDGE BSs are currently defined only for GPRS data services. Thus, the EDGE BSs will be interconnected only to the GPRS network nodes with a Gb interface. To support voice services on the BSS associated with the new modulations, two possible options are foreseen:

Build a connection from the BSS to the IS-41 network infrastructure.
Enhance the GPRS system to make it more suitable to carry real-time services like voice.

The first option will likely require less feasibility study. For example, it can be implemented with a traditional A interface concept. The second option is expected to be more challenging, since potentially a new paradigm like voice over IP is introduced.
Voice over IP is part of the family of applications driving the need for QoS support. These applications are the same as in the wireline domain, namely, interactive real-time applications such as voice over IP, videoconferencing, and, to a lesser degree, other real-time applications such as streamed video on demand, remote learning, telemedicine, and Motion Picture Expert Group (MPEG)-4-based applications. GPRS and IP QoS support, both of which are in their infancy but maturing quickly, is crucial to materialize the aforementioned applications. However, the uncertainty lies in the dimension of mobility, whose impact on the network QoS provisioning is hardly well understood in packet-switched networking. In order to support services that require QoS, the GPRS QoS classes have to be mapped suitably onto IP QoS mechanisms in order to support QoS in the entire GPRS PLMN. In addition, the admission control procedure used for computing the RLC/MAC layer priority parameter has to be extended to the IP backbone as well for the same reason. Finally, the nodes making up the IP backbone in GPRS must implement appropriate scheduling and queuing mechanisms to support appropriate forwarding behavior. After all these interworking issues are sorted out, the effect of mobility has to be studied and understood.
Another class of services based on the distribution or many-to-many communication model will evolve in the cellular sector. IP multicast is a protocol used on the Internet for both real-time and non-real-time applications requiring one-to-many communication. Based on this protocol, many services such as stock and news feed, live video broadcast, and streamed pay-per view services may be provisioned. The key issues concerning multicast-based services are layered QoS provisioning for heterogeneous users and application adaptation.
MPEG-4 is an emerging technology for future multimedia and may be a driver [5]. Although in the United States MPEG-4 is currently only considered an important software technology for Internet applications, in Japan it has been spun as a technical element critical to upcoming third-generation W-CDMA mobile phone. MPEG-4 applications being explored include portable videophones and terminals, pocket worldwide Web connectors, and downloadable video multimedia applications.
MPEG-4 supports the coding of images and video with scalability and error robustness for mobile communications. Scalability includes spatial, temporal, quality, and content scalability to enable the decoder to only decode a part of a bitstream and reconstruct images or video sequences. The scalability is desired for video transmission over heterogeneous networks such as mobile networks, as well as for applications where the receiver (e.g., mobile terminal) is not capable of decoding or displaying full-resolution or full-quality images or video sequences. Furthermore, for mobile communications in error-prone environments, MPEG-4 provides error robustness and resilience to allow transmission of image or video information via wireless networks. Therefore, GPRS-136 systems should consider MPEG-4 applications in their evolution.

Conclusions

The adoption of GPRS as the packet data solution for IS-136 provides most of the advantages of regular GPRS, namely mobility management coordination between circuit switched and packet switched systems, GSM economies of scale and migration roadmap to 3G, and global data roaming. Services enabled are not just existing IP services, but new services rendered possible by the efficient, tetherless, and ubiquitous access to packet data networks, coupled with the user's mobility.
The SGSN tunneling principle, invented to minimize changes to both IS-136 and GPRS systems, allows one to connect GPRS network nodes to non-GSM systems without requiring the GPRS nodes to interpret non-GSM protocols. Thus, a great deal of flexibility is achieved for future evolution.
System evolution beyond GPRS-136 will take place in two dimensions. The first is the radio interface; EDGE is already planned as a new modulation for 136 that will achieve even higher bit rates for packet data services. Beyond EDGE, W-TDMA and W-CDMA are possible, but spectrum availability will be an issue to be addressed. The second dimension is the network architecture, which will have to evolve to support new services, especially multimedia.

References
[1] CDPD Sys. Spec., rel. 1.1.
[2] "Digital cellular telecommunications system (Phase 2+). General Packet Radio Service (GPRS); Service description; Stage 12, "GSM 02.60 v. 6.1.0, EN 301 344, rel. 97.
[3] TR45 TIA/EIA-136-336 Draft Text v. 1.4, Dec. 8, 1998.
[4] S. Blake et al., "An Architecture for Differentiated Services," IETF draft.
[5] Overview of the MPEG-4 standard.

Additional Readings
[1] Wireless Application Protocol Architecture Specification.
[2] "Digital cellular telecommunications system (Phase 2+). General Packet Radio Service (GPRS); Service description; Stage 1," GSM 02.60 v. 6.0.0.

Biographies
Stefano Faccin is a senior research engineer at Nokia Research Center, Irving, Texas. His research interests include development of novel architectures for third-generation mobile networks and beyond, and the study of security infrastructures for mobile communication systems. Prior to joining Nokia Research Center in March 1998, he was a research engineer with CSELT (Centro Studi e Laboratory Telecomuni-cazioni), Italy. He has an M.S. degree in computer science and telecommunications from Politecnico di Torino, Italy.
Liangchi Hsu is a senior research engineer at Nokia Research Center, Irving, Texas. From 1983 to 1986 he worked with Industrial Technology Research Institute, Taiwan. From 1989 to 1993 he was a senior engineer with Tandy Electronics R&D Division, Fort Worth, Texas. He worked for Motorola, Fort Worth, Texas, as a staff engineer from 1993 to 1997. He received a B.S.E.E. degree from National Chiao-Tung University, Taiwan, in 1981, and M.S. and Ph.D. degrees in electrical engineering from the University of Texas at Arlington in 1988 and 1996, respectively.
Rajeev Koodli is a senior research engineer at Nokia Research Center, Burlington, Massachusetts. His research interests include IP QoS, IP multicast, and congestion-friendly transport protocol design for continuous media applications. Currently he is working on mobile network design and implementation. Prior to joining Nokia Research Center in August 1997, he was a graduate student at the Electrical and Computer Engineering department of the University of Massachusetts, Amherst, from which he has M.S. and Ph.D. degrees.
Khiem Le is research manager, Mobile Networks, at Nokia Research Center, Irving, Texas. The work in his group includes the investigation of GPRS adaptation to non-GSM systems, and of novel network architectures for future mobile networks. His relevant experience includes management of various PCS and cellular systems engineering projects for the regional Bell operating companies at Bellcore. He also acted as a consultant to various companies like TRW on the Odyssey project and Sprint PCS. He has a Ph.D. degree in computer engineering from the University of Southern California.
Rene Purnadi is a research engineer at Nokia Research Center, Irving, Texas. His research interest is developing network architecture to support next-generation wireless mobile telecommunication. His current work includes developing call control and mobility management in the wireless network and air interface. Before joining Nokia Research Center, he worked for Motorola. He graduated from the University of Texas at Austin, where he earned his Ph.D. in numerical analysis.