PAPER: THE EVOLUTION OF GSM DATA TOWARDS UMTS Kevin Holley, Mobile Systems Design
Manager, BT, United Kingdom [Editor’s
note: Abstract GSM data services have launched a new era of mobile communications. The early analogue cellular modems were unattractive to the market as they were slow and unreliable. GSM now offers a high quality 9600 bits per second link which opens the way for useful data applications and the Short Message Service provides guaranteed delivery of small data packets even if the phone is switched off when the message is first sent. Now the market for data is moving onwards (more bursty) and upwards (more traffic), and the ETSI standards groups are working towards higher data rates but more significantly also towards packet data services, initially with GPRS and multislot data but eventually UMTS will provide much higher bandwidth.. This will certainly broaden the appeal to end users because data is routed more efficiently through the network and hence at lower cost, and also access times are reduced. Yet there are still very few applications which make the most of mobile data. Time is wasted through the use of inefficient user protocols and it is rare for an application to provide fast recovery after call failure. It is very important for application designers to consider the capabilities of the lower level bearer and this paper aims to identify some of the key aspects of this task.
Cellular access to data services has been around for a long time. Even in the mid 1980s engineers were busy designing modems which could counter the effects of radio fading. Yet the data access services were not popular. Connections were unreliable and slow. Equipment was bulky, with the term "soap on a rope" in vogue as users struggled with many boxes to connect between the phone and the data terminal. GSM has turned this around, with high quality data solutions involving reliable connections and compact phones with only a single wire between phone and terminal. At the same time as GSM networks became popular, the growth of the Internet started to reach out to the general public. The now-ubiquitous web browser has inspired a generation of data users and it is possible to find nearly everything somewhere on "the web". Large companies all over the world have embraced this technology and made it their own by providing in-house Intranets, firewalled from the outside world yet accessible to all in the company.
Figure 1 - Businesses are very dependent upon corporate LANs The mobile computer user has also seen dramatic changes over the last 10 years. The "laptop" computer is now virtually extinct, with high speed colour PCs now available. Much smaller handheld devices called Personal Digital Assistants or PDAs have emerged and become very popular. With these developments in the role and presentation of data particularly in businesses the main demand for data services can be described as open access to public information on the Internet, or secure access to private information on company Intranets by a mobile population using handheld PDAs or notebook PCs. But this is only what the end user wants to see, so how can we achieve that in reality?
GSM was conceived to be the "mobile arm" of ISDN, using signalling based on Q.931 and providing access to all the ISDN data services. The communications path is digital so the speech is converted to ones and zeros before transmission.. This made it quite easy to introduce data services because there was no need for any modem at the user end. GSM is a widely adopted standard with roaming (enabling subscribers from one operator’s network to communicate via another operator’s network), a key consideration of all developments. Thus it is possible to use the same application in many countries without the need to reprogram the dialled number or any other setup change. GSM provides two basic data services, transparent and non-transparent. The transparent service provides a consistent delay and a throughput of 9600 bits per second, with variable error rate according to the propagation conditions. The non-transparent service provides a consistent low error rate with a throughput up to 9600 bits per second and variable delay according to the propagation conditions. Whichever "bearer type" is chosen, there are many options for connection between the GSM network and the far end data network. The most commonly used ones are PSTN modem and ISDN data service. With either of these, the throughput is limited to 9600 bits per second and so flow control and/or rate adaptation are required. The rate adaptation as well as the basic data conversion from GSM protocols to the fixed network protocols is performed by an Interworking Function (IWF).
For the mass-market, the most obvious data service to offer is the non-transparent service, as this provides built-in error correction. However we have already seen devices on the market which allow end to end data compression by using the transparent data service. Claims of 36000 bits per second can be met and in some cases exceeded, depending on the type of data being transmitted. In addition GSM has now developed a standard data compression capability built on V.42bis, which can be applied between the mobile terminal and the network. This service will allow non-compressed data from a users terminal to be compressed in the mobile station and de-compressed in the network, thus overcoming the limitation that the reduced speed on the air interface gives. Facsimile devices are very complex. They change modem speed during the call, have training to determine the acceptable data rate, cannot be flow controlled and sometimes even operate to a non standard protocol. This makes it very difficult to insert a digital bearer in the middle of a fax call. To get facsimile to work with GSM, the standards-makers had to terminate the fax protocol and re-create it, both at the terminal side and in the interworking function. Despite the complexity, GSM provides a very good facsimile group 3 capability. The GSM standard incorporated a two-way paging function. Called Short Message Service, or SMS, this allows 160 characters of text to be sent to or from a GSM handset. Messages are sent to a Short Message Service Centre (SMSC), which stores SMS until it can be successfully delivered to the destination. Whilst in many cases the first attempt is successful, it may be that the destination handset is switched off or out of coverage, so the SMSC makes many attempts to send the message, possibly over several days. SMS has two distinct service classes: Mobile Terminated (messages sent to the mobile) and Mobile Originated (messages sent from the mobile).
There is another short text-based messaging service called Cell Broadcast or CBS, which has the ability to broadcast messages in a given geographical area, making it suitable for information type applications available to all subscribers in a given area.
All of these data facilities have been implemented for some years by the manufacturers. We have seen the emergence of the GSM data card, a PC card which looks like a modem to a PC, but instead of converting the PC data to modem tones it converts the PC data to a proprietary protocol which is then in turn converted by the handset into standard GSM data protocols. The GSM standards-makers, however, always envisaged that GSM handsets would provide a direct data connector, and this has been achieved by some manufacturers already. Some rely on additional PC software drivers but others effectively implement the entire GSM data stack in the mobile phone. These PC interconnect solutions provide for data and fax calls to and from the mobile user. The Short Message Service / Mobile Terminated is now implemented in all mobile phones, and quite a few also have the Mobile Originated capability too. However mobile phone keypads are limited in terms of the speed of data entry and a PC interconnect solution for SMS makes the business of entering and sending messages much easier. The PDA market is not served as well by the GSM handset makers. PDAs have difficulty in supporting a PC card slot because the power requirements are significant. They are also limited in terms of memory and CPU power and so complex communications drivers are a non-starter. The best solutions for PDAs are provided by manufactures who have a phone with direct data connectors (physical or other e.g. optical). Some manufacturers have taken the step of integrating PDAs with phones. These are attractive because everything comes in one box and all services work together seamlessly. However, the offerings to date have been quite limited because they are large for a phone and do not have all the capability which people expect from a PDA. In the future we expect much more from this area.
The general trend is for data applications to generate increasingly bursty data streams, this makes for inefficient use of a circuit switched connection. With present data traffic levels accounting for only a small part of the mobile network usage, there seems little need to worry about efficiency. However, fixed networks have seen an enormous growth in data traffic, not least because of the rise and rise of Internet access demand, and there is no reason to suppose that this will not spread to the mobile networks as technology and customer expectations move on.
The existing GSM network is optimised for circuit switched voice calls and therefore is inefficient at setting up and carrying very short data bursts.. Figure 4 shows a comparison made of the number of radio timeslots required to carry data traffic generated as the user base increases. It shows that with only a few subscribers the same number of channels are required for circuit and packet switching. As the user base (and traffic volume) increase then the number of channels required increase faster if circuit switching is used. Packet switching is more efficient at carrying bursty data, and mobile networks should employ packet switching right up to the end terminal if they are to cope with the expected demand for future data services. The network and radio capacity required to support a large amount of bursty data would make it uneconomic or impossible (ie limitations on the number of physical radio channels available) to implement.
A packet switched mobile network opens the way for a range of new applications. IP interconnectivity and Point-to-Multipoint (PTM) transfers become a reality with GPRS, applications currently restricted to fixed line connections will migrate to the mobile world.
The very robust 9600 GSM data service works with good quality of service throughout existing GSM cells. With any design for high speed mobile data there is a trade-off between data rate and coverage area, and GSM has provided a good data speed with coverage wherever there is voice coverage. A strong requirement for a new faster service has emerged with a more relaxed requirement for total coverage. Thus the GSM standard will soon include a 14.4k bits/s data service as a basic enhancement to the existing bearer service classes. This will be achieved by reducing the error protection mechanism and therefore it will not be possible to use 14.4k bits/s over the whole of the GSM service area, instead requiring the user to stay in good to moderate coverage. As well as increasing the rate for a normal call, the GSM standards will offer timeslot concatenation. To explain this it is necessary to go into a little more detail about the GSM radio interface. A single radio carrier on GSM is 200kHz wide and provides up to 8 simultaneous channels. Each logical channel is given one eighth of the time on the physical carrier. But a single call could potentially be allocated more than one of these logical channels and hence the aggregate data rate could be higher. The GSM multislot standard will combine up to 6 channels to give the user a rate up to 64k bits/s. There is no benefit in increasing the data rate beyond this because the switch network is based on narrowband ISDN 64kbits/s circuits, so the rate limitation starts to be the core network rather than the access.
But the biggest change is that of providing packet radio access. The new Generalised Packet Radio Service (GPRS) will do just that, it will offer operators the ability to charge by the packet, support data transfer across a high speed network and up to 8 timeslot radio interface capacity. GPRS introduces two new nodes into the GSM network - the Serving GPRS Support Node (SGSN) and Gateway GPRS Support Node (GGSN). The job of the SGSN is to keep track of the location of the mobile within its service area and to send and receive packets from the mobile, passing them on, or receiving them from the GGSN. The GGSN then converts the GSM packets into other packet protocols (IP or X.25 for example) and sends them out into another network.
With GPRS, a terminal logs on to the network separately for GPRS. It then establishes a Packet Data Protocol (PDP) context which is a logical connection between the mobile and a Gateway GPRS Support Node. Having done this it is now visible to the outside fixed network, e.g. the Internet and can send and receive packets. There are four radio channel coding schemes defined for GPRS, to allow the data rate to be increased when coverage is good, and ensure packet loss is never too great when coverage is not so good. These coding schemes allow between 9.05kbit/s and 21.4kbit/s per GSM timeslot in use. Thus the maximum data rate for GPRS is 8 times 21.4kbit/s or approximately 164kbit/s. The performance of all GSM data services will vary according to radio conditions. However it is possible to categorise GSM data services according to a few key elements: * access time - how long it takes to get into a ready-to-transfer-data state * throughput - users data transfer rate through the GSM system * packet size - Size of data that is managed as one entity through the GSM system
Shown graphically, this gives a much better idea of how the limitations of narrowband ISDN switching impact on the maximum throughput when comparing GPRS with multislot data (see figure 6).
Figure 6 - GSM Data Service Maximum Throughput
All data services can be broken down into a few key components as follows:
All of the GSM data services can be used with any of the interconnect options in the centre column, however clearly the bit rate will be very important to Video in particular. The main element in choosing the GSM service is the size of "packet". For very small packets sent unicast or small-scale multicast, then SMS would provide an ideal transfer mechanism over GSM, whatever the end interconnect requirements. For larger packets then GPRS becomes the most attractive bearer, but this will not be available for a while so circuit switched data may have to be used. The terminal in use will also be important, not only from its size but also how much data is required to be stored. A despatch service would only need a few messages, whereas an email application would require more storage. Whilst this way of studying the requirements is very useful in terms of defining the service to use, it is of paramount importance to consider the service required by the user rather than looking at the technology available and making mobile access possible. The following examples show what a difference this can make to the user perception:
The difficulty with moving forwards from the present position is that there is a need for more than the traditional systems integration tasks. As well as providing user software and drivers, the systems integrators need to get additional functionality built into the Office Automation, Email and Internet Service Provider business so that online times can be reduced. There is some scope for additional value added service businesses in taking the Internet data and converting to a more mobile-friendly format before passing on to the mobile user, however for the Corporate customer this is unlikely to be popular in the long term as concerns over the security of intranet data are very high. Services like multislot data and GRPS are very useful in moving the base technology forwards, but if the same goals can be achieved with the existing data services, we should be looking at prototyping services now on the current networks. We need to produce reliable user solutions which can cope with dropped calls without restarting a session, perhaps automatically re-establishing should this be needed. We should avoid using inefficient protocols for mobile just because they are existing and in common usage. People need access to common services, not necessarily using the same network protocols as are used in the higher bit-rate, lower error-rate fixed networks. The time is now right to look at producing a standardised mobile access mechanism for fixed network services, focusing on increasing the effective throughput and immunity to dropped calls and thus reducing the needed airtime.
There are many goals yet to be achieved with wireless data. The whole data industry is moving away from X.25 based services towards TCP/IP based services. Data transport is becoming cheaper and cheaper as economies of scale drive down the cost of routing equipment. User devices are becoming more sophisticated and smaller. The idea of email and web browsing in your pocket can be achieved today yet it is expensive and in some cases awkward to use. But the mass market for data devices is the key to future success. Defining that mass market and the most useful services is the key. We must make the best use of the available data services to avoid excess cost. The long-term goal is the UMTS vision, quoted from the UMTS Forum Regulatory report as: "The Universal Mobile Telecommunications System, UMTS, will take the personal communications user into the Information Society of the 21st century. It will deliver advanced information directly to people and provide them with access to new and innovative services. It will offer mobile personalised communications to the mass market regardless of location, network or terminal used." Basically the implications of this are that all telecommunications networks are integrated and users really can get fast data rates of 2Mb/s in some areas. But 2Mb/s is only useful for the right applications and those applications need some market experience before they can be well defined. There are bound to be some winners and some losers but at least some of the battles must be fought with the existing second generation cellular capability of up to around 64kbits/s. The future expectation of growth in the mobile market is driven by data and multi-media. This expectation is now being realised on fixed networks, mobile networks through the enhancement on GSM and introduction of new capabilities such as GPRS will start to build that market. UMTS is the "utopia" for meeting the market expectation, however, it needs to be technologically advanced sufficiently to meet that need.
GSM data provides a wealth of opportunity for transferring a variety of information to the user on the move. From simple email to video and multimedia, almost anything is possible. The future will bring even higher bandwidths, larger markets and with them lower costs. However in order to achieve the goal of mass market data use the applications must be developed with mobile in mind. Users will not want to bother with redialling just because of a dropped call, and they will not be happy to pay for two timeslots when one would suffice if only the data was pre-compressed and used an efficient transfer protocol. We need to look at the system as a whole, starting with the users needs and ending with the data transport mechanism. Terminals will be very important to the success of mobile data. Whilst there will be roles for all types of data terminal, the most successful ones will provide unified mailboxes, not one button for fax, another for internet email, yet another for X.400 and a fourth for SMS. The future is bright for data, if we get it right!
Acknowledgements The authors would like to acknowledge the support provided by colleagues at BT Laboratories, in particular Alan Clapton and Chris Fenton. Biographies Kevin Holley is currently working as a Mobile Systems Design Manager with BT. He has been involved in the development of the GSM standard since 1988, particularly on the data services and Short Message Service. He also works closely with the UK operator Cellnet in defining their GSM services. He is now chairman of SMG4, the ETSI data development group for GSM and UMTS. Tim Costello is currently working as a Cellular Systems Engineer with BT. He is working on future mobile data applications and particularly GPRS. |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
