Re: The problems of congestion in mobile networks in Nigeria
By Balarabe Sani
It is widely acknowledged that there are serious issues of quality of service on Nigerian telecommunications networks and the NCC has taken up the challenge by daily interfacing with the operators on these issues.
The Commission has also recently introduced new key performance indicators (KPIs) to measure the quality of service offered by them and is sanctioning those who fail to meet the quality threshold.
However, it is important to carefully analyse the write-up by Professor Augustine Odinma which has appeared severally in some Nigerian newspapers.
The publication by Professor Odinma on congestion in mobile networks is subjective, unprofessional and baseless because of lack of the use of conventional techniques of testing quality of service (QoS) on mobile networks. First of all, in any statistical method as the professor knows, the data sample size is very important in ascertaining the correct result.
The more we increase the number of data samples, the more correct the statistical result. Although the professor claimed to have enough samples for the evaluation, we will explain why the figures quoted were erroneous.
Wrong Methodology of Measurement
Professor Odinma claimed that about 72,334 calls were made during the test within a six-day period from three fixed locations in Lagos, Yola and Abuja. That was an average of 12,000 calls a day.
If we study the modern network performance testing methods, we find out it will not be possible to make 12,000 calls in a day in the manual testing method claimed by the professor. Remember that automated equipment test several GSM networks at the same time in one time slot.
A manual testing can only test one handset at a time and therefore would be impractical to have up to 12,000 samples per day even if the test will be extended to 24 hours.
Lack of Considering Mobility Management Function of Mobile Network
Professor Odinma and his co-researchers had left or forgotten one major aspect of mobile technology in their experimental research — mobility management.
Mobility management is an important phenomenon that distinguishes mobile technology from fixed network. The network has the capability of seamlessly handing over an ongoing call by a mobile user to a neighbouring cell in a process called Handover.
If signal strength degrades to some level in the serving cell, six neighboring cells are tested by the network and the one with better signal strength is selected. A stationary user as claimed by the Professor in his research does not consider the mobility aspect of the mobile network.
A stationary user as in the case of the professor’s testing methodology may experience congestion because he is hooked to only one cell, called serving cell that may have weaker signal or traffic congestion at some instant.
The professor and his co-researchers’ testing of only single cell (one each) from three different locations does not test the performance of mobile networks as a whole, because large number (thousands cells precisely) of cells make the complete network. For the professor’s test to be meaningful he must move to test as many more cells as possible.
Another thing the professor failed to take into consideration is the bouncing busy hour on a mobile network. Unlike in a fixed telephone network, in a mobile network the busy hour is not fixed, it is dynamic over time.
For example the mobile user who lives in Victoria Island (VI) in Lagos may experience high congestion during working hours and less congestion in the evening and weekends.
This is because in the evening and weekends more people have moved out of VI. The people who live near market places may also experience similar phenomenon.
Failure by the professor to consider the bouncing busy hour phenomenon in a mobile network environment by conducting test from three fixed locations, will lead to erroneous results since mobile environment was regarded in the experiment to be similar to a fixed network environment. Lack of inclusion of the mobility management aspect of the mobile network technology in his test leads us to believe that the Professor has a limited knowledge of the mobile network QoS testing methodology.
Flaws in Professor’s QoS testing Methodology
One important parameter in mobile QoS testing is “Mean Holding Time” of a call. The professor did not tell us the mean holding time he and his co-researcher used in their trial calls.
If holding time is very long the result will never be good because of call drops. If holding time is too short the result may also not be good because of signaling requirement for setting up of calls.
The Professor mentioned that they initiated 35,816 calls from Lagos location within six days using ten handsets. This represents an average of 5,969 calls per day and an average of 597 calls per handset per day (from Lagos location).
In mobile QoS testing, it is expected that calls are made during peak usage hours.Let us assume that the two research students stationed in Lagos by the professor started placing their test calls at 7:00 AM, and continue testing up to 10:00 PM. In the night after 10:00 PM up to 7:00 AM next morning no call is expected because of light traffic. This represent total of 15 hours of manually testing calls.
It is internationally accepted for network planning that the Mean Holding Time for a call is 90 seconds (1.5 minutes), in Nigeria it might be slightly lower due to flashing. The 15 hours of testing calls by the two students represent 15 x 60 or 900 minutes.
Bearing in our mind that average call per handset per day from Lagos location is 597; assuming that the Professor used internationally accepted mean holding time of 90 seconds and neglecting all the overheads for setting up of call ( time wasted during trials, post dialing delay, number of ring tones before the receiver picks the call, call clearing time e.t.c.).
The length of time spent on call with one handset is 597 x 1.5 or 895.5 minutes. It can be deduced that Professor Odinma’s claim of initiating 35,816 calls with ten handsets within six days from Lagos location by only two people is impracticable because this figure is close to total of 15 hours (or 900 minutes) of continuous call with one handset without resting. We should also remember that there are ten handsets to be used by only two people in this manner, unless each person holds five handsets and used them to place calls one after another without resting, eating, drinking or visiting wash rooms for 15 hours per day.
It should also be noted that the handset must not be on-continuous call, because it should be half of the time idle to respond to calls from other remote locations. Even if the Professor used much smaller mean holding time than the internationally accepted standard, his result is still invalid because his test did not represent an ideal user pattern.
Conventional testing Techniques of QoS Measurement on Mobile Networks
Conventional methods of QoS measurements on mobile networks objectively are two. The first is the call data acquired by the network performance system in the Operator’s Network Management Center (NMC). Millions of calls data acquired by the network performance system at the NMC can be processed to give accurate QoS measurement statistics of the network within the period of interest (hourly, daily, weekly or monthly). Large sample size of number of calls data acquired in this method is adequate to generate accurate QoS statistical result.
Also data for millions of calls acquired at the Point of Interconnection (PoI) can give accurate statistics of PoI performance. The acquired data at PoI includes incoming and outgoing traffic from one network to another, Answer Seizure Ratio (ASR) and utilization of interconnect circuits in all the interconnect routes.
The second method is through a “Drive Test”. In this method a vehicle is equipped with equipment that automatically places calls on all networks or at least networks of interest. The equipment is so sophisticated that user pattern can be programmed.
The user pattern includes a number of trials at a time, time pause between another trial and mean holding time of the call, etc. The mobility aspect of the mobile network is fully captured in this case, because the vehicle moves with the equipment within small or large area of interest, city, state or even region and can acquire call data (from many cells) for days to weeks. Data acquired during the drive test can be processed in the post processing center.
All valuable QoS statistics can be generated for intra and inter-network calls. This method is also accurate because it fully considers the mobility aspect of the network and large number of calls is tested covering a wider geographical region during the drive process. In addition to hundreds of QoS statistical parameters the equipment can perfectly evaluate speech quality of connected calls.
Professor Odinma should understand that the methodology adopted and the locations chosen were erroneous in manual testing of optimization in GSM network and will never lead to any meaningful result.
The professor should also understand that in order to get the best performance of mobile technology and probably measure quality the user should not be stationary in one location (i.e. served by only one cell). In this scenario the mobility management function is jeopardized.
The neglected handover process by the Professor in his research was unfortunate because handover is very important in mobile networks, through handover process the network can seamlessly handover call to a better neighboring cell that may have more capacity and better signal quality to handle the call.
Some popular equipment for measuring QoS on mobile networks are QVoice and TEMS. The Professor can log onto www.ascom.com, the QVoice equipment manufacturer’s Web site (who also acquired TEMS from Ericsson last year) in order to educate himself with internationally agreed techniques for measuring QoS on mobile networks. We advise that, next time the Professor wishes to undertake a similar research, he is advised to either use any of the above mentioned equipment or collect network performance data from the operators (acquired at NMCs), since the data is acquired as a result of millions of calls information and it covers wider geographical location.
- Eng Dr Balarabe Mohammed Sani, is the Director, Technical Standards and Network integrity, Nigerian Communications Commission, NCC.