re: Issues with GPRS (EMC) Part 2 - long but comprehensive
>> Many Obstacles Lie Ahead As Operators Aim For GPRS Roll-Out, Conference Hears
Dennis Andrews
EMC Cellular
11-Jun-2001
e-searchwireless.com
Network Capacity Issues and GPRS Limitations
Although high usage of GPRS does have implications for overall network capacity, most operators seem to be confident that the growth rate will not exceed their current capacity growth plans. While user numbers are low, operators will be reluctant to allocate timeslots for exclusive GPRS use. Under these circumstances, voice traffic will take precedence over packet data and users will notice a reduced response time. It also appears that where the network is unable, for any reason, to establish a GPRS session in response to a request from a user, most handsets will default to dial-up mode; and this may not be apparent to the user unless there is close inspection of the display icons. This clearly has billing implications and may result in unexpected high charges being presented to the user.
Where there is strong demand for higher speed data service, operators have, in theory, the option of implementing the higher rate coding schemes 3 & 4. But these would require extensive and costly upgrades to infrastructure and the general feeling is that any significant deployment is unlikely.
Taeke Castelain, from BlueKite, explored some of the issues that would arise from an unexpected, rapid commercial uptake of GPRS service. With a current practical limit of 4-timeslots due to radiation and heating issues and relatively high network latency (the overall roundtime response for an application), there should be an understanding of the 'transport efficiency' for different types of traffic - ie the associated potential value versus volume. Values are compared, taking values for a typical speech application as a base. This shows that the relative efficiencies associated with proposed GPRS and 3G applications quickly falls; and this will be aggravated with the inclusion of colour support and enhanced graphics - both considered essential for success in wireless data markets. Castelain pointed to the advantages that will come from advanced methods of data compression that are marketed by his and other organisations.
Content Volume (kbyte) Value Transport Efficiency
Speech 100 (1 minute) 1 1
Message/Text 1 (1 message) 0.5 50
Web browsing 1000 (5 minutes) 1 0.1
Images 1 (1 image) 1 1
Audio 1000 (1 song) 0.5 0.05
Video clip 1000 (30 seconds) 5 0.5
Video call 2000 (1 minute) 3 0.15
This theme was followed up in a number of presentations from operators that have already launched service. T-Mobil summarised the steps required to maximise performance:
Step 1 - Optimise throughput and latency of transmission chain
Step 2 - Adjust TCP parameters to the GPRS case (MTU, window size etc) in the proxy between the internet server and GGSN
Step 3 - Speed the proxy by
- Usage of recent HTTP version (e.g. HTTP 1.1; request pipelining, persistent connections)
- Lossless compression (e.g. gzip on HTTP layer)
- Lossy compression (maximum jpeg, reduce number of colours/pixels)
- Caching
- On average, acceleration by a factor of two is achievable without dramatic loss of picture quality.
Handsets, Power Management, Chipset Development
A comprehensive report on handset availability was published by EMC at the time of CeBIT at the end of March 2001 () and there was little further information available at the Congress. There were suggestions that manufacturers are holding back on delivery of new models while interoperability issues are resolved. Some suppliers, Siemens being mentioned as one, are planning to allow field upgrades of handsets to allow for revisions to operating protocols that might be introduced to overcome early implementation problems. There was a wide view that a high proportion of handsets will eventually support GPRS as a standard feature.
Richard Lodge, from Tropian, reviewed the issues affecting chipset development for 2.5 and 3G systems.
High data rates depend upon use of multiple timeslots. GPRS Class 12 operation can demand up to four timeslots in either direction. The resulting increase in power amplifier current, together with increased CMOS dissipation due to the higher computation load, potentially causes a large increase in thermal dissipation requirements of the handset. If EDGE is implemented, the PA efficiency drops to less than half of that realised by the current GMSK modulation method. This will result in further heat dissipation requirements
Manufacturing issues. There are a number of manufacturing challenges concerned with power control accuracy and repeatability, phase error variation, switching transients repeatability, and parametric variability.
Chipset designers are developing new technologies which will address these issues for both 2.5 and 3G. Amongst these new techniques is Polar Impact Technology which is a digital, strongly non-linear transmitter solution, which as well as reducing overall power consumption and consequent heating effects, significantly reduces component count in the device.
'Always-on' Issues and IP Addressing
During the conference, the issue of GPRS being an 'always-on' system was questioned on a number of occasions. It appears that, at least in the short term, there will be many occasions when a GPRS terminal is not 'always-on'.
Issues arise from a number of factors.
* IP addressing - there are just not enough IP addresses available for an individual designation to be issued to each GPRS handset; this means that connected terminals must use a dynamic IP address assigned by the network for the duration of a session. It follows that where the session terminates, the handset will, effectively, no longer be connected. This will appear to the user as a delay in response as the handset waits for the network to re-establish addressing. An alternative approach for operators is to employ Network Address Translation (NAT) where the network operates a system of 'private' IP addressing and where address translation takes place when access to the global internet is required
* Network congestion - an 'idle' GPRS terminal will still consume network resource even though there is no exchange of data. This may become a significant overhead for operators if numbers of users grow and they may need to take steps to limit this for low revenue users
* Handset issues - it appears that some of the early handsets (even so-called Class B handsets) do not have the in-built functionality to maintain network sessions when non-data features are used.
Interoperability
Operators and manufacturers are, naturally, somewhat reluctant to provide much detail over the technical issues that might inhibit a smooth roll-out of GPRS service. However, from comments made during various presentations at the Congress, it is clear that most of these arise from the variations of interpretation possible in the SMG31 version of the GPRS standards. These variations of implementation lead to problems with hand-off between cells and problems in timeslot allocation. Both of these difficulties lead to problems in either the establishment of a GPRS session or a dropped call.
There was criticism voiced that there are few published testing scenarios through which operators can prove operational efficiency of various handsets against different networks. In practice, operators have been forced to restrict operation to handset units that, by experience, have been shown to operate satisfactorily.
Additional interoperability issues arise from differences in WAP browser support, although these were not specifically mentioned in detail during the Congress.
Roaming
GPRS roaming was highlighted by a number of speakers as being of prime importance, especially if business users are to be attracted to the service. Although a small number of inter-network roaming connections are already in place, these are generally private arrangements between individual operators of the same group and cannot be compared with the more or less global roaming arrangements that apply to voice service. The technology that needs to be added to networks to allow data roaming is relatively straightforward; but it is the organisational and contractual issues that appear to be delaying implementation.
In its simplest form, a network can allow data access to a visitor by setting up a local IP connection, providing it can obtain call authorisation and billing information from the home network. However, this would provide a presentation that is linked to the local network data service and would conflict with an overall desire that the user sees a common application presentation regardless of whether he is operating on a home or visited network. This will be especially important where WAP is a primary application interface as the roamer will require access to his home network WAP application set.
The additions required to allow full roaming are a Border Gateway (BG) that provides a secure roaming link out of the network, and a range of Domain Network Servers (DNS) that resolve IP address requirements. The issue holding up development of GPRS roaming revolves around the way in which the networks will then link their border gateways together in order to exchange data.
There are three potential methods of achieving this linkage.
1. Through the public internet. This would be a least cost solution but would be inherently insecure and would not provide a predictable level of service
2. Leased lines. Each operator could maintain a network of leased lines providing direct connection with every other operator with which it has established a GPRS roaming agreement
3. Through a VPN specially constructed and maintained for handling GPRS roaming.
The initial GPRS roaming services have all made use of dedicated leased lines. However, future development is certain to be based on the VPN solution; but the timescale for widespread implementation is uncertain. The GSM Association's International Roaming Expert Group (IREG) has established a set of standards known as Global Roaming Exchange (GRX) which defines protocols to be used by operators and independent 'GRX providers'. Sonera was one of the first organisations to trial GRX functionality and there are a number of others that have announced similar plans. But, at present, there is no real indication of when a viable and widespread service offering will become available.
Other topics
Tariffing, Prepay, and Billing
Only a few presentations dealt directly with revenue recovery.
Operators that have already launched GPRS service have a relatively wide range of tariffs in terms of periodic charges and fees for data quantity. However, with a lack of experience to reflect the amounts of data that users might transfer, the included data quantities are usually quite high; this means that these tariffs are essentially flat rate rather than the supposed volume based. Some operators have instituted tariff schemes that give a choice of WAP-only, or WAP+WEB access.
Ilan Paretsky, Director of Marketing at Comverse, looked at the requirements for developing advanced systems that will provide a range of tariff and billing options, for composite voice and data users, that will include the facility for content-based billing - ie charges associated with the inherent value of the supplied data. It was accepted that there will be consumer resistance to paying for internet derived information where the user has been accustomed to receiving this 'free' in the fixed environment. Although the Camel-3 protocols will support complex billing scenarios, they will not include any element of content value billing and this will need to be supported by independent servers.
With an increasing proportion of cellular users opting for prepay options, there is a lot of interest in developing billing mechanisms that will extend GPRS availability in this area. This will be particularly important if the concept is accepted that the 15-24 year-old age group - big SMS users and predominantly prepay - are to be convinced of the attraction of migrating to GPRS.
Developments in North America
Two presentations, from AT&T, and Golden Bridge Technology/SBC dealt with the evolution of GSM-based wireless data in North America. Despite recent suggestions to the contrary, both speakers were at pains to emphasise a continuing role for EDGE as part of their development strategy towards 3G. Both gave timescales that saw a GSM/GPRS overlay to their respective networks during 2001, followed by an upgrade to EDGE during 2002 and the beginning of UMTS deployment during 2003.
Experience of early implementers
Operator presentations came from:
Amena (Spain)
Blu (Italy)
Omnitel (Lithuania)
T-Mobil (Germany)
Telering (Austria)
Telstra (Australia).
....but little was revealed that was not already in the public domain.
Presentations were limited to reiterations of the basic marketing concepts and detail of standard technical implementation. In response to questions, the speakers were extremely reluctant to talk about GPRS subscriber numbers beyond admitting that early take-up was slow and hinting at numbers '...in the low thousands'.
Acceptance of Advertising
A number of speakers discussed possible user attitudes towards advertising and the potential for this separate source of operator revenue to allow a lower tariff scale for specific groups. There were clear views on the constraints that would need to apply.
* Any advertising must be personalised and relevant to an individual users interests
* It must be optional with the ability to opt-in and opt-out at will
Advertising must be unobtrusive - text and graphics would be acceptable, voice messages unacceptable
* Quantities should be moderate - say around three per day, or 10% of SMS messages
* Advertising must always be free to the recipient and not paid for through wasted airtime.
M-commerce
Although presumably of significant importance to the ongoing development of data services, there was virtually no mention, during the Congress, of m-commerce issues. It would appear that these are currently very much on the back-burner until the lessons of the recent problems in the dot-com world become apparent. <<
- Eric - |
