This section gives a detailed overview of how wireless data services are provided in the most popular cellular environments. Similar to the previous chapter, we have broken up the information according to first-, second-, and third-generation systems.
Cellular and PCS circuit-switched data (CSD) technologies have been around for over a decade. VPN techniques can be utilized with most of these systems. 1G wireless technologies such as TACS, NMT, and AMPS, described in Chapter 3, are based on frequency-division multiplexing. A voice signal is transmitted directly on a chunk of spectrum that is dedicated to the user. This same chunk of spectrum can be used to modulate data signals using a wireless modem, similar to what happens with the 3-kHz band of telephone copper pair used for modem dial-up data connection. Error correction protocols used by wireless modems tend to be more robust than their landline counterparts because of the necessity to deal with a more challenging physical environment with inherently higher interference and signal-to-noise ratios than copper or fiber cable.
The data rate for an AMPS modem call under good conditions is as high as 14.4 Kbps and under poor conditions is as low as 4.8 Kbps. There may be a delay of 20 or more seconds to establish a connection. As a result, the user experience is not entirely satisfactory, and combined with the poor price-performance ratio, this has made the CSD services over first-generation cellular systems an insignificant source of revenue for carriers. This partially explains the certain stagnation of data services during the 1990s relative to voice in the middle of a heyday of 1G cellular systems.
Circuit-switched data has not vanished with the first-generation systems. Instead, it made a logical transition and is now available as an option in both 2G and even upcoming 3G systems. Second- and third-generation systems are based on digital transmission of speech over dedicated radio resources (either TDMA time slots or a code in CDMA, see Chapter 3). These systems are based on the use of digital channels, and therefore modulation and demodulation do not need to occur in order to transmit data from the mobile stations to the core network. These systems have been augmented with data circuit-switched services capabilities and, as a result, may offer transparent data service or "reliable" service by using Automatic Repeat Request (ARQ) mechanisms, which involve retransmission of lost chunks of data transmitted over the digital channel. Also, these channels need to be terminated at some place in the wireless network either to provide direct access to a data network or to interwork with ISDN or PSTN lines, in order to provide end-to-end connectivity to an IWF in the wireless carrier's core network or a router in a network owned by another party, such as a corporation or an ISP. In the following sections, we explore the capabilities offered by GSM, UMTS, CDMA (IS95), CDMA2000, and TDMA (ANSI-136).
Circuit-switched data service in both North American TDMA and CDMA IS-95 systems is provided similarly to other 2G cellular systems. Shown in Figure 4.2, this service requires a fixed circuit—in this case, a CDMA modem dial-up connection—to be established between the mobile and the call destination. The system must carry the data in a digital form to an IWF, which can then generate the modem tones for communication through the PSTN.
In fact, the users who subscribed to a circuit data service offered by a wireless carrier can simply use their cell phones as terminals extending physical layer to connect to external modems in IWF, often supported within the MSC. An IWF is effectively splitting the mobile data call into land and mobile segments. The 2G mobile phone, and in fact the radio network (including the cell tower, base station, and MSC), acts as a conduit for passing commands to the PSTN-connected modems in the IWF. In a sense, this system provides a virtual serial cable (similar to that which would connect a computer and an external modem on the desktop in a landline remote access scenario) that extends from the mobile computer to the modem in the IWF.
To this effect, a data call originates at the IWF in service provider networks, which dials out to the mobile's destination number. IWF converts CDMA or TDMA -encoded data into normal analog modem format and passes the call over the PSTN. CDMA airlink supports two data rates: 9.6 Kbps and 14.4 Kbps. Additional compression techniques often allow users to achieve data rates up to 19.2 Kbps. The call sequence for a typical CDMA circuit data connection includes the following steps, which are illustrated in Figure 4.2:
The circuit connection is established by the MS.
The data call is directed by the MSC to the IWF.
The outbound call is placed by the IWF modem over the internal IWF packet bus to a TDM switch.
The TDM switch connects the call through the PSTN.
The call is terminated by the modem at the call destination.
One of the popular commercial enhancements to cdmaOne data system is Quick Net Connect (QNC).  This technology, originally designed by 3COM, Qualcomm, and Unwired Planet (now renamed to OpenWave systems), to allow carriers to provide users with direct access to the Internet and private IP networks, completely bypassing the PSTN and avoiding the need to use modem dial-out procedure. In QNC the IWF routes data from the CDMA wireless network to the Internet or intranet using a PPTP or L2TP (see Chapter 2 for more information on these tunneling techniques). This avoids the use of PSTN network resources, PSTN setup time, and modem training time, resulting in faster and more reliable end-to-end connection.
GSM was originally defined to offer digital speech services, since it was designed at the time when the transition to digital speech was recognized as a necessary improvement to the existing 1G systems promising to increase their capacity, voice quality, and confidentiality. GSM deployment was also considered an essential step toward easier international roaming. Although initially the GSM system did not encompass data access, as computer network access became increasingly important to customers as a new service, and to operators as a new source of revenue, new system capabilities were defined. At this time connection-oriented CSD services as well as connectionless data services such as SMS and Unstructured Supplementary Services Data (USSD—which now sometimes is used to interact with Wireless Application Protocol, or WAP, gateways) were defined.
In GSM, the data call is most often initiated by the GSM subscriber, who establishes a dial-up connection by placing a call to a desired data access number provided by an ISP or a private network. Initially the data call originated by the MS is terminated at an IWF, which then converts the incoming data into regular analog modem format and conducts a dial-up procedure to establish a link over the PSTN to the final point-of-call destination, that is, a remote access server (RAS). PPP normally provides the end-to-end link layer service between the MS and the RAS.
The data received by the IWF is converted to the analog frequency-shift keying (FSK) tones characteristic of analog modems. The resulting end-to-end data call can be viewed as two independent calls: the mobile data path and the PSTN path. The mobile data path refers to the connection between the mobile and the IWF. The PSTN path refers to the connection between the IWF and the remote access server at the call destination.
An alternative to this architecture is the Direct Internet Access feature, which is similar to CDMA QNC. With this feature the PPP link that originated at the mobile is terminated at the IWF and then IP packets are routed directly to the Internet. On the downside, the mobile user is limited to the Internet access and other data services provided only by the wireless carrier. This drawback, however, can be avoided by relaying the PPP frames received by the IWF over L2TP, thus enabling the wireless carrier to provide access service to remote networks, by using an L2TP VPN-based approach.
GSM circuit-switched users are offered approximately 9.6 Kbps or 14.4 Kbps of throughput. Actual throughput normally depends on the radio channel conditions and the possible compression protocol negotiated at PPP session setup.
With regard to CSD there is no architectural difference between UMTS and GSM, other than the UMTS ability to provide higher data rate CSD services through the use of more efficient radio interface technology. GSM in turn defined High-Speed Circuit-Switched Data (HSCSD),  which we discussed in Chapter 3 and whose capability is described in [GSM TS 02.34]. HSCSD allows for the concurrent use of multiple full-rate traffic channels (TCH/F is a 16-Kbps channel) to achieve greater data rates. The HSCSD MS can also request asymmetric bearers.
An MS could request transparent transfer of data over the wireless network circuits. In this case the GSM/UMTS network is simply providing the user with a channel that delivers bits from the MS to the external network, and from the external network to the MS. This kind of service is normally not appropriate for the transfer of TCP/IP-based applications data because of the relatively high bit error rate that would affect communication to the point where TCP/IP throughput would be very low and, at times, close to zero. Instead, this service may be useful for unrestricted delivery of bit streams.
The GSM and UMTS CS domain offers a set of basic bearer services listed below [GSM TS 02.02]:
UDI. This service provides transfer of unrestricted digital information.
3.1 kHz. Service used to select a 3.1 kHz audio interworking function at the MSC, this service category is used when interworking with the ISDN or PSTN 3.1-kHz audio service and includes the capability to select a modem at the interworking function. "External to the PLMN" indicates that the 3.1 kHz audio service is only used outside of the PLMN, in the ISDN/PSTN. The connection within the PLMN, user access point to the interworking function, is an unrestricted digital connection.
PAD. Through this service, access to a PAD (for X.25-based PSPDN) is provided.
Packet. This service provides direct interworking to a packet or ISDN network (normally X.25, X.31, and so on).
Alternate speech and data. This service makes possible swapping between voice and data during the duration of the call.
Speech followed by data. With this feature, it is possible to start a call in speech mode and then to switch to data, but not the other way around.
Of these, we will be looking only at the UDI and 3.1-kHz services. In most cases, the service is based on the 3.1-kHz bearer, since we are considering PPP-based access for the purposes of advanced data services. UDI and 3.1-kHz service are the most widely used at this time, so we will focus on them.
3.1-kHz and UDI bearer services are very much related. In fact, the 3.1 kHz term does not apply to the PLMN, rather it represents the PSTN resources used to provide end-to-end connectivity (the spectrum required to transmit a voice conversation) that a modem at an IWF can use to connect to the data network accessed by the PLMN subscribers. Typically UDI and 3.1 kHz use the same physical bearer from mobile to IWF, and different bearers after IWF toward the external (to the PLMN) network. UDI is normally used to support ISDN-type connectivity, while 3.1 kHz is used for analog connectivity to the external networks. Typically, V.110- and V.120-compliant terminal adaptation is used for ISDN connectivity. Using UDI with V.110 or V.120 is also possible to provide direct digital access to data networks, thus achieving faster connection setup.
UDI and 3.1 kHz services are characterized by a very important QoS parameter partially covered earlier in the chapter: the transparent or non-transparent types of the bearer service. As mentioned, a service is transparent if the network behaves simply as a bit pipe, totally unaware of what user protocol is being transmitted. In nontransparent mode the MS and the IWF are the termination point of the Radio Link Protocol. This protocol provides a set of services to the user protocol relay functionality. The user protocol (PPP for common network access services) is being encapsulated in RLP frames. These frames are normally numbered and smaller than the user protocol, so that the retransmission unit size is normally between one and two orders of magnitude smaller than the original user protocol packet or frame size.
This allows retransmission of fewer octets over the air when a part of the user data gets corrupted because of bad radio conditions. The numbered frames can get retransmitted—via ARQ—and also sequence numbers can be used to implement flow control mechanisms, so that when there is excess incoming traffic that the radio interface cannot handle, the RLP sends backpressure signals to the relay functionality (at the terminal side and at the IWF), so that user traffic flow can be paced appropriately. The RLP enhances the link quality from a bit error rate (BER) point of view; that is, it provides some degree of independence of the BER from the radio fading conditions. Though this independence may have some throughput costs, it is normally better than what a transparent service not attempting retransmissions over the radio interface would achieve.
The rest of this chapter focuses on packet-switched data, since we believe it is more likely to be found in future networks and will provide more efficient foundation for delivery of advanced data services such as Mobile VPN.
Quick Net Connect is only available for the call originating from the mobile device.
Note that HSCSD is also defined in UMTS framework.