Cellular Networking Perspectives

Protocols Layers: Can there ever be too many?
Protocol layers are, to a computer or communications engineer, like mothers and apple pie. Who can be against them? Every modern computer protocol is layered, and all play at least lip service to the OSI 7-layer model (Physical, Datalink, Network, Transport, Session, Presentation and Application). The ideal of protocol designers is to focus on only one layer at a time, ignoring layers above and below. Paradoxically, however, this approach has resulted in useful protocols either failing to achieve market success or being delayed by several years in implementation.

Protocol layers are generally designed in a horizontal fashion, like an archaeologist who concentrates on one layer of history at a time. However, protocols are always used in a vertical fashion. Only by piling up the right layers can a working protocol stack be developed. Not only must a workable stack be defined, but anyone that you want to talk to has to use a compatible stack. It is an emphasis on the horizontal at the expense of the vertical that has caused many problems with protocols.

IS-41 is, today, a highly successful protocol, used around the world to provide national and even international call delivery, handoff and other roaming services. Yet it took much longer to get widespread implementations than it should have, and even today international roaming is seriously limited, all due to protocol stack problems. The IS-41 application layer is well defined, but the transport layer can be either X.25 or SS7. Beyond that, X.25 could be used in a level 2 (point-to-point) or level 3 (packet switching) configuration – and SS7 could be used with only the MTP layer or with both the MTP and SCCP layer. Four different configurations to run one application! And, if the two companies attempting to connect made a different choice, they would not be able to communicate without protocol conversion, even though their IS-41 applications were 100% compatible. I was involved in a field trial where first two, and then three, of the four possible transport protocols were used simultaneously. Needless to say, the trial had to be abandoned. Eventually, the industry settled on two major protocols (X.25 level 2 and SS7 MTP), and now uses X.25/SS7 protocol conversion equipment where necessary. Perhaps in the end it is a better system for all the confusion, but the extra flexibility resulted in at least a two-year delay in the widespread implementation of IS-41 services.

Today, IS-41 faces another challenge, with increasing interest from carriers in Asia, Australasia and South America. SS7 has become the de facto standard network protocol for IS-41, with X.25 relegated to a lesser role in network access. Yet the version of SS7 used is ANSI (American National Standards Institute), used in the US and Canada, not the ITU (CCITT) version of SS7 used in the rest of the world. What’s an international carrier supposed to do?

For now, the solution is to extend the US cellular signaling network to any country that wants to support international roaming using IS-41. This means these countries have to run two parallel SS7 networks: an ANSI network to support IS-41 and an ITU network for all other uses. The ideal solution is to use international SS7 gateways to convert from ANSI to ITU, but this is not possible with only the MTP layer in use (which allows for national addressing, not international). An agreement is needed on how to configure the SCCP routing layer – a piece of the vertical puzzle that has not yet been solved. SS7 is perhaps unique in the flexibility of its addressing capabilities. Through the definition of a new “Global Title,” new addresses can (theoretically) be added at will. But, this flexibility comes with a price. All companies have to either agree to support multiple addressing schemes or agree to compromise on a single scheme. And not only the players that wa nt to communicate must get involved, but the companies that control the STPs (Signaling Transfer Points that provide global title routing services) must also agree to configure and manage the new routing tables.

A protocol that has suffered tremendously from a lack of depth has been IS-124, a very rich application layer for the exchange of call detail information related to wireless systems. Its capabilities hold great promise for enhanced fraud and billing applications, along with a world of applications for marketing, customer service and network monitoring that have yet to be dreamed up. Yet, for all its promise, implementations of IS-124 are still in the very early stages. A large part of the problem is the lack of definition of a complete set of vertical layers. This allows companies to implement to the specification without being able to communicate. Eventually, this will be resolved when the industry finishes converging on a common protocol stack, but at least four years will have been lost since its initial publication in 1993.

The lack of success of wireless data is one of the mysteries of our time. It has certainly suffered from the lack of common vertical solutions. Perhaps the most successful form of wireless data is the use of an analog modem on an analog cellular channel. It is so successful because it will run any protocol that has been adapted to standard modems. Consequently, a huge variety of packaged vertical protocol stacks are available. Physically, these stacks are visible to the consumer as the floppy disks that Compuserve, AOL and internet service providers hand out like candy. Each disk contains a stack of protocols that allow access to their services using virtually any modem. Other forms of data transmission (RAM, ARDIS, CDPD, digital cellular, etc.) often require the development, configuration or acquisition of some part of the vertical stack by the user. Too often, users of new data protocols have to struggle to accomplish connection to a modem, which then results in achievin g only what they could have achieved with less hassle by using a modem with analog cellular to start with.

The internet may prove to be the savior of wireless data protocols. It provides a rich set of protocols for email, file transfer and browsing that are not reliant on modem-based access (although certainly compatible with it). Wireless data designers should latch onto this set of protocols because it allows their users to obtain a complete vertical stack while they only have to design the access layer. In protocol design, applied laziness is very important.

It is both essential for new data protocols to live within a well-defined vertical stack, yet it is impossible for any organization to define the complete stack. A major component of good protocol design is to define the existing building blocks that users of a hot new protocol will want it to rest on and those that it will support. Without this insight, and the application specific knowledge to make the right analysis, many important and needed new communications protocols will continue to underachieve.