Not so fast!
Ian Channing
01 April 2001
Far from being the smooth transition promised in the late 1990s, GPRS is in fact proving to
be hard going for vendors and operators. In fact, the concerted effort to get GPRS to
market quickly could be creating as many problems as it solves
Uncertainty over handset availability is usually quoted as being the reason for the
continuing delay in commercial launches of GPRS. The complexity of GPRS, however,
means that technical issues regarding the network may be making their own contribution
to the delays in GPRS commercialisation.
The industry as a whole appears to have underestimated the complexity of GPRS. The
common view was that GPRS required the addition of new network elements and some
software modifications but was, in essence, an upgrade of the existing GSM infrastructure.
This attitude failed to grasp the problems inherent in overlaying a packet-switched network
on to an existing, and very busy, circuit-switched network. It also failed to grasp that this
was the first move into the altogether different world of IP. "It is very important to optimise
the IP parameters because TCP/IP, which is being run over the GPRS connection, has all
sorts of timers and parameters," points out Russell Whitworth, managing consultant at
telecommunications outsourcing company WFI. "The GPRS standard does not say
anything about TCP/IP but if you want to make efficient use of the network you need to get
things like packet sizing right. There is a lot of tuning that people are only now finding out
how to do that will improve the throughput. If the IP parameters are wrong then you will see
less than ideal throughput under any circumstances."
Another major issue impacting on GPRS implementation is the considerable amount of
flexibility within the standards. Of necessity, standardisation bodies cannot address every
single possible option; there has to be flexibility for vendors to develop their own solutions
for particular situations.
The GSM specification offered this flexibility, for example with the A interface, but the
GPRS standard has many more options than GSM.
Flexibility
To aid early market entrance, vendors are using the flexibility within the standard to create
proprietary solutions across large areas of GPRS - which in turn is creating interoperability
problems. "There are two things that happen with standards," comments Whitworth. "One
is that standards typically leave you with some choices in implementation and it takes
time to iron those out. The other thing is that standards sometimes leave areas open that
people have not thought about. It is only when you come to build the network that you
realise this and then manufacturers make different choices. This happened with WAP and
with GSM in the early days and it took time to thrash out."
The rush to get to market also means that handset and infrastructure suppliers are not all
singing from the same hymn sheet as Steve Baker, GPRS product manager at TTPCom, a
developer of technology and intellectual property for wireless communication terminals,
explains: "The specifications did not really stabilise until the middle of last year. We were
ready to do [GPRS] testing in December 1999, then we discovered that the infrastructure
vendors had changed to a different revision of the specifications...In fact, until the middle of
last year, we did not even have a common view of the platform."
Andy Bond, senior product manager at Actix, which offers optimisation solutions for mobile
networks, outlines the problems that arise when not only are there multiple revisions of the
standard in the market, but vendors are pushing to get their products out: "Six or nine
months ago, the infrastructure was on revision SMG28/29 but the handsets are at SMG27.
The vendors are actually coding and trying to deliver equipment right behind the
specifications and also trying to deploy equipment at the same time. This means you end
up with everything being contracted into a very small space in time in order to meet the
roll-out targets."
"People underestimated the complexity of GPRS and, because the vendors are naturally
competing with one another, they are all trying to get into the market early and this means
that they are having to develop proprietary solutions whilst they are waiting for the
specifications to catch up.
Where there is a gap it will be filled one way or another and it is guaranteed that the
vendors will not fill it the same way."
One of the major concerns of operators planning to offer GPRS services is how to achieve
quality of service. Although the industry has pulled back from earlier extravagant claims of
data throughput speeds in excess of 100kbit/s, there was still confidence that significant
data speeds, perhaps up to 40-50kbit/s, could be offered in due course. Given certain
limitations within the GPRS technology, such speeds could not be guaranteed to every
user for every transaction, the majority of customers having to settle for best available.
However, part of the long-term strategy of operators was to be able to offer guaranteed high
and constant data speeds to customers who were prepared to pay premium prices.
Quality of service
According to Raymond Wu, head of product management at Ascom, which offers network
assessment, analysis and control equipment, there is still a lot of work to be done to
realise this happy state as there is no way that operators can offer quality of service
guarantees over today's GPRS networks.
"There is no mechanism technically today to actually guarantee QoS in the air interface. It
is written in the standard that you have parity, precedence, packet delay and so on, but
there is no way of actually implementing in the air interface today. Guaranteeing packet
delay or mean throughput for any subscriber is a real challenge."
In fact, says Wu, the situation as it stands is that as a GPRS user you are always be
competing for capacity with voice users "and you will always lose. If you are in a
congested area and someone comes in with a voice call when you are transferring data
over a GPRS link, then your data throughput will be throttled back and in extreme cases
your data transfer may stop completely until some capacity is again available." One good
way in which this problem can be overcome is for operators to allocate specific timeslots
in areas such as city centres exclusively for GPRS traffic. The Catch-22 of such a strategy
is that by doing so, particularly at this early stage of GPRS development, the operator will
lose voice revenues. Wu says he knows of no operator which has yet taken the step of
committing capacity specifically for GPRS. "If you don't allocate capacity for GPRS you
will have the situation where sometimes it is there and sometimes it is not - and when it
disappears it takes a long time to come back."
Even if the problem of QoS in the air interface is solved in the longer term, the operator's
ability to offer a reliable and constant service is still in doubt. As Wu points out, once the
call has left the GGSN, the quality of service is beyond the operator's control: "Once you
get into the world of IP you don't know who is sharing the resources with you. Everyone is
excited by the shiny new world of IP, but if you mention QoS to anyone in the IP world
they will nod their head in sympathy. In going from circuit-switched to IP we are facing the
same problems as they have with the internet; all that GPRS does is get you into that
cloud marked 'internet'. If the internet can't solve its QoS problems then GPRS is not going
to help. In fact, GPRS poses its own problems on top of those of the internet."
Testing GPRS
These are also frustrating times for the test and measurement industry.
Initially the lack of GPRS handsets made air interface testing impossible; now some
handsets are available but connecting to the GPRS network in a consistent manner is
proving difficult. Some vendors even report that, during testing, the GPRS network
completely disappears - without explanation - and fails to reappear for considerable
periods. Denton Clutterbuck, business development manager EMEA, Wireless Network
Solutions at Agilent Technologies, says that although handsets are now available,
connecting to the network is still a difficult process: "We seem to have overcome most of
the technical issues with regard to the handsets but for some time we were still unable to
connect to the network. This focused the problems more on the infrastructure than on the
actual terminals. Now we have reached the point where nine times out of ten we can
actually connect to the GPRS network with our test equipment, but there are still
difficulties. These are difficult to pin down and are due to combinations of phones,
infrastructure, coding schemes used and the quality of service set up within the operator's
network and in the handset. There are still quite significant technical challenges to
overcome, not so much on the phone but on the infrastructure before we will see a
successful launch." This complexity does not bode well for the launch of commercial
GPRS networks. Even though most operators are planning to introduce the service to
selected corporate customers, if connecting to GPRS proves time consuming for skilled
engineers, how difficult will it be for untrained users?
Interoperability
There also continues to be issues regarding the interoperability of terminals and base
stations from different vendors. WFI's Whitworth says that it is still early days and there
are "a lot of teething troubles involving compatibility between the handsets and the
networks. These are going to take time to iron out." Whitworth goes on to point out that
this situation also prevailed in the early days of GSM but was satisfactorily resolved.
Of equal concern is the question of the differing technical status of the handsets and the
base stations. Infrastructure vendors are deploying base stations which can deliver all the
basic features of GPRS but which do not, because of the need to get the product out,
include some of the more esoteric GPRS functions. Handset manufacturers can evaluate
their products against the current functionality of the base station but are unable to check
whether it will continue to work when the base station is upgraded at a later date. As Steve
Baker of TTPCom points out, this places both operators and handset vendors in a difficult
position: "Fundamentally the handset manufacturers are not going to ship a product that
cannot be tested because the infrastructure vendors have not finished fully developing their
products. The vendors don't have to demonstrate every capability that is in the
specification in their first shipments. So there is a sword of Damocles hanging over the
terminal manufacturer because if he starts shipping a product which is only tested against
a partial implementation of the network, what is going to happen when the network is fully
implemented?
Does he go to product recall? Are operators going to accept that they are going to have to
give new phones to subscribers when they upgrade their networks?"
Right first time
In addition to the problems already enumerated, there are also issues concerning network
optimisation, the type of services that operators should offer and the volumes of traffic they
will generate, equipment availability - the list goes on.
Nonetheless, getting GPRS right first time is vital. Following the PR disaster of WAP and
the lack of confidence in the wireless industry currently being felt by the world financial
community, another failure cannot be contemplated. Also, GPRS has a vital role to play
both in bridging the growing gap before UMTS and also in developing the market for
wireless data services. Even if it takes longer than expected, it is essential that the
technology is right when commercial GPRS services come on stream. |
