At most locations the panel visited in Europe and Japan, it saw a need to design an architecture for wireless networks that was capable of providing much higher data rates than exist in second generation cellular systems, regardless of the details of the air interface standards. Furthermore, the need exists for multimedia delivery to a small (handheld or laptop) terminal. At the present time, the focus is in supplying World Wide Web (WWW) and general Internet services. While there is limited potential for doing this using the PSTN, which the second generation relies on, such an approach is likely to prove inadequate and both the air interface and the core network switching and routing will need to change to accomplish the task. There are a number of competing approaches. The two most prominent are (1) wireless ATM and (2) some form of packet radio. Both of these ultimately envision packet or cell transmission end to end. That is, including the air interface, but also fundamentally changing the routing and switching in the core network away from traditional voice oriented circuit switching. Among the many variations of these proposals, the one that stands out because it is currently implemented and shows promise for smooth evolution to the full convergence is General Packet Radio Service (GPRS).
GPRS is an outgrowth of GSM/TDMA in Europe. Its basic architecture is the transmission of packets and is designed to support IP while minimizing hardware modifications of existing network elements. The packet format and its derivation from the GSM eight time slot air interface are show in Fig. 3.4 using slot number 4.
Fig. 3.4. GPRS packet format.
This can provide about 18 kbps of true throughput (overhead excluded) per slot. Several slots can be used to increase this rate. The layered architecture for GPS is shown on figures 3.5a, 3.5b, and 3.5c. Here is illustrated the interface, Um, between the mobile transceiver and the Base Station Controller (BSC).
Fig. 3.5a. GPRS architecture.
Fig. 3.5b. GPRS network.
Fig. 3.5c. GPRS routing.
Once the packets get to this point, they can be switched and routed by any kind of packet switching network such as the Internet. The next step in the evolution of GPRS is Enhanced GPRS (EGPRS). This provides for a higher radio interface rate and more flexible user rates. Enhanced Data Rates for Global Evolution (EDGE) will be introduced to boost network capacity and increase the data rates of both circuit switching using High Speed Circuit Switched Data (HSCSD) and packet switching (GPRS) up to three fold. Possible rates may then exceed 400 kbps. The evolution of GPRS is believed by many to be the migration path towards a Universal Mobile Telecommunication System (UMTS) and the UMTS Radio Access Network (UTRAN)
An alternative contender for providing integrated wireless access and core network switching and routing is Asynchronous Transfer Mode (ATM) or perhaps more properly called "cell switching." Originally developed for wireline multimedia services, ATM has connection oriented features and the ability for differentiated services and negotiated bandwidth plus Quality of Service (QoS) guarantees, that has much to recommend it. One can build an entire end to end network based on ATM. Wireless ATM has become a candidate architecture that assumes that multimedia service will be ATM based. Figure 3.6 illustrates the basic architecture. In this structure, developed by NEC Research, the air interface is ATM cells and a special ATM access point is provided to serve each geographical microcell. A new Mobile Network to Network Interface (M-UNI/UNI) is added. The User to Network (UNI) is standard. If IP networking is required then the proposal suggests carrying IP packets over ATM (IP/ATM).
Fig. 3.6. Network architecture: Wireless ATM system.
Wireless ATM appears to have some serious shortcomings. For example, since Adaptation Layer 2 (AAL2) appears to be the ideal mode to support both the radio interface and the core switching network, the use of minicell packets (up to 42 octets) within AAL2 requires specific overhead for signaling and switching, the result is a significant loss in efficiency. In addition, for high mobility users the Wireless ATM scheme may not be easily scalable.
GPRS has its own problems in that it is a gradual approach to achieving the convergence objectives and in the initial stages it is still tied to 2nd generation wireless techniques for roaming and mobility. Furthermore, once the packet gets into the TCP/IP network, it is not distinguished from non-wireless traffic. Radio transmission is notorious for its many impairments such as fading, multipath, shadowing, and blocking which causes packets to be dropped or corrupted. The existing TCP protocols interpret this as congestion and take action that can severely reduce performance. In addition, TCP/IP is a connectionless service with no QoS guarantees and until new differentiated service features are introduced, a completely satisfactory solution is not possible. Other problems exist in the interaction of the higher layers of the protocol stack because of the anomalies of the physical layer that exists in wireless, and especially in high mobility and handoffs.
A possibly fruitful area of research for switching and routing in wireless networks is packet radio taken to its limits. Packet radio per se has been investigated for many years as an application in military tactical communications.
The ultimate solution may be to design the wireless network with an advanced version of packet radio more suitable for the global commercial market. The large geographic coverage requires some form of cellular structure where the mobiles act as relays to a cell site (or satellite) for long haul. In addition, the structure of the network is very "fluid" so that "ad-hoc" networks may be established without prior configuration. There is considerable research already underway for small scale ad-hoc wireless networks but that needs to be extended.
There is also a need for understanding the interaction of the higher layers of the protocol stack when they are being serviced by an unreliable and quirky wireless physical layer and the need for handoffs, mobility, and roaming. A further positive note for direct wireless packet networks is that there are new protocol suites such as Ipv6, which have a very large address space, built in security features and characteristics that increase efficiency and performance. Another new direction in Internet development is Multiprotocol Label Switching (MPLS), which bears directly on the issues of the ability to independently route and switch many connections destined for the same address and allows a unique ability to optimize traffic flows and to emulate connection-oriented virtual circuit switching. MPLS has been advanced by the Internet Engineering Task Force (IETF). It is a label switching technique that integrates layer 2 switching with level 3 routing. Label switched routers can improve performance and provide for differentiated services and multiple protocols, including Ipv4 and Ipv6, among others. Although these ideas are being developed primarily with the wired network, they have many potential benefits for the new multimedia broadband wireless systems.