UTILISATION OF MOBILE NETWORK PERFORMANCE
DATA FOR DYNAMIC CAPACITY REALLOCATION
Jaakko Rissanen VTT Information Technology P.O Box 12031 FIN-02044 VTT, Finland email: [email protected] ABSTRACT
Traffic load in current mobile networks increases all the time. The performance of the networks can be optimised by analysing the performance data of the network, exploiting location information provided by the network or the mobile terminals and utilising adaptive antennas. The process includes detecting the areas that have capacity problems, locating the area where more capacity should be added and making the relevant changes to the network configuration. This paper addresses the automation of mobile network capacity reallocation and introduces an application for realising it.
KEYWORDS
mobility, location-aided planning, capacity reallocation.
1. INTRODUCTION
The number of users of mobile cellular networks has increased rapidly in recent years. The load of the networks varies in different areas and times, exceeding network capacity occasionally. To be able to adopt to the changing traffic conditions in the network both geographically and temporally, the operator has to be able to adjust the network elements, especially antennas, according to analysis carried out on actual performance data of the network.
The Adaptive Coverage System (ACS) concept developed in the EU IST project Cellular Network Optimisation based on Mobile Location (CELLO) (Cello Consortium 2002) is the idea of exploiting information about the location of mobile terminals for adjusting the antennas of a mobile network in order to achieve increased network capacity and stability. The changes in the network are done according to predefined network plans which describe the cell configurations to be used in the network. ACS has been proven to provide means for changing the network configuration in a feasible manner. By switching the antenna lobes to desirable directions, the system is able to enhance the network capacity in certain hot-spot areas. In a trial carried out in summer 2002, it was also shown that this can be done in a well-controlled and predictable manner in an operative network (Jormakka et al. 2003). To develop the concept further and make the changes in the network configuration based on the actual performance of the network, dynamic network capacity reallocation has been studied.
This paper is based on the results of the CELLO project and it discusses different aspects of dynamic network capacity reallocation. The realisation of the concept in CELLO is done by introducing a new module called Data Analyser Module (DAM), which
is described in Section 3. Although only GSM network is considered in the current implementation, the approach can be employed also in other mobile systems.
2. OBTAINING PERFORMANCE DATA
Forming a realistic view of the performance of a mobile network can be done by analysing data from the Operations and Maintenance Centre (OMC) of the network (3GPP 2002). OMC collects a vast amount of data concerning traffic channels, control channels, communications between various network elements, calls etc.
In order to utilise the data in the most optimal way, Mobile Network Geographic Information System (MGIS) has been implemented in the CELLO project. The MGIS system is capable of collecting and storing location-related performance data from mobile terminals and the network including OMC data. The data stored in MGIS can be used for detecting problem areas in the network and therefore it provides means for location-aided network planning. For capacity reallocation, the most interesting information is obtained from the cell level data, that is the data concerning operation of unique cells in the network. In addition, the Terminal Level Measurement Samples (TLMS) included in the MGIS database can also be used for determining the load of the network. These samples are formed from specific measurements containing information about the serving and neighbouring cells, especially received signal levels. In addition, for these samples there exists the exact location from where the sample has been collected. It can be used for locating the precise hot-spot area.
3. ANALYSING THE CAPACITY PROBLEM
There are three main problems that dynamic network capacity reallocation should find solutions to. Each of them considers different aspect of the topic and together they form the process of the analysis needed for the optimisation of network capacity. The problems are listed below and they are described in detail in following sections.
• Detecting the cells with capacity problems in the network. • Locating the area where more capacity is needed more precisely.
• Defining the changes to be made in the network configuration in order to add capacity to the problem area.
3.1 Detecting the Cells with Capacity Problems in the Network
The detection can be done by analysing OMC data of the network or Terminal Level Measurement Samples collected in measurements. Because lack of capacity cannot be always discovered by examining only one performance parameter of a network, an approach using three parameters obtained from performance data was chosen.
The most obvious parameter would be the blocking rate in traffic or control channels, but that would mean that the problem had already occurred because OMC data is usually updated on an hourly basis. In addition, the behaviour of a GSM network in a congestion situation is such that the first indicator of upcoming congestion is the high utilisation of
signalling and traffic channels. Only after that the corresponding blocking rates start rising (Kyriazakos et al. 2002). Thus the first performance parameter to analyse is the traffic load in the network cells. In addition to OMC data, Terminal Level Measurement Samples can also be used for defining the load of the network by calculating them, taking into account the place and time of the samples and defining a threshold which indicates high traffic.
If the planning of the network coverage has not been done carefully enough, undesired coverage problems may appear in certain areas. This can be noticed by examining handovers and their causes. In addition, at areas with large amount of network traffic, the number of handovers tends to be high because of the interference caused by other users of the network in the same area. Thus the amount and the causes of handovers form the second parameter in the analysis.
In specific situations such as mass events or emergencies, the amount of users of the network may become very high in a small area in a short period of time and thus more network capacity is needed to serve all the call requests. To demonstrate this, the third parameter to be analysed is based on the detection of a specific message received from the network, namely an emergency notification.
Altogether, these three parameters are combined into one equation, which is used for the calculation of value K for each cell. This value describes the traffic situation of that cell. It is used for deciding which cells' areas need more capacity. Calculation of the value K for each base station BTSi is defined as
( )
( )
∑ = ∗ + ∗ + ∗ = n j i BTS j HO c i BTS T b i BTS E a i K 1where a, b, and c are weighting coefficients and E( ), T( ) and HOj( ) functions that
calculate the amount of emergency messages, traffic load and handovers due to causes j respectively.
In order to allocate the resources of the network optimally, the system should analyse performance data constantly. When value K exceeds a predefined threshold in any of the cells of the network, an event report is generated and sent for further processing.
3.2 Locating the Problem Area More Precisely
Once the need for more capacity on cell level has been detected, the next task is to find out a more precise location where capacity should be increased. The information that has been gathered in the previous phase tells only that more capacity is needed in the coverage area of a cell. Of course, the exact location of the Base Transceiver Station (BTS) which serves the cell is always known. For the capacity reallocation point of view, this is not the optimal location because the location of the mobile users is usually other than that of the base station. Thus, a more elaborate reasoning is needed.
By using OMC data only, there is no other choice than the above-mentioned BTS location. With the MGIS database there are more possibilities. The information about the antennas in the network can be used. As a rule of thumb, a good guess would be to estimate the average location of the congestion to be 500 meters to the direction of the main lobe of the antenna. Information about the radiation pattern of the antennas in the network can be found from MGIS database. However, as the shape and size of the cells'
coverage areas are not uniform and they are always shaped by topography, this can sometimes lead to incorrect results. Therefore, additional information is required.
The coverage area files of the network cells can be used in the estimation. They are also included in the MGIS. These files contain the received signal level values in the coverage area of the cell estimated by the network planning tool. By using these values as weights, the centre of gravity of the coverage area can be calculated. This is a reasonable estimation of the place where the congestion could occur and more capacity in the form of coverage from another cell should be directed to.
When using Terminal Level Measurement Samples in the analysis, the area with capacity problems can be located even more precisely because of the available location data of the measurement samples. During the time period of the analysis, the measured samples' locations are accumulated. When a capacity problem has been noticed, the average of the sample locations is calculated. In this calculation, the samples collected most recently must be weighted in such a way that their influence is larger than that of those collected in the beginning of the time period. This leads to the most accurate estimation of the location where more capacity is needed.
After locating the problem area, the actual changes to the network configuration must be defined.
3.3 Defining the Changes to be Made in the Network Configuration
In order to add capacity to the defined problem area, the network configuration has to be changed. The Adaptive Coverage System being implemented in the CELLO project is using Modular Antenna Arrays (MAA) (Eggers et al. 2001). MAA is a phase-steered antenna which can be dynamically focused and steered to a hot-spot area by flexibly setting the antenna radiation pattern. There are several possible narrow beams and one wide beam available. In practice, the wide beam covers the area of all the narrow ones. One of these beams can be in use at a time. An example of the different beams is presented in Figure 1. In the analysis, the challenge is to find the correct MAA setting in order to increase the capacity of the hot-spot area in the most efficient fashion.
The outcome of the previous phase of the analysis is the location of a hot-spot area. The next task is to detect which of the available, predefined network plans excluding the currently used one adds the most capacity to that location. The network plans include all the combinations of MAA settings that have been defined.
The network may consist of both traditional, fixed antennas and adaptive antennas such as MAAs. Since the radiation patterns and directions of the fixed antennas are static, they are excluded from the analysis for performance reasons. Only the MAA cell configurations and their respective coverage files are examined.
The MAA cell configurations whose coverage areas do not include the hot-spot location are discarded. For the remaining ones, the coverage area with the highest predicted signal level at the hot-spot location is selected. Because in reality the capacity problem never occurs at a single, exact location, a slightly larger area can be used. For example, signal level values within a circle of certain radius can be averaged and the result be compared with that of other MAAs.
Once the coverage area and the corresponding cell configuration have been found, the network plan containing this cell configuration is identified. If the resulting network plan is
Figure 1. Polar plot of the beams of a phase-steered antenna
not the one currently in use, the analysis has been successfully performed and the network plan is returned as a result. Otherwise, the analysis returns a notification that the best possible network plan is already in use and no further actions are required.
Since the changes in the network should not decrease the grade of service experienced by users connected to other cells of the network, the current load situation at the cell whose antenna would be directed to another location must be checked. If the cell, and possibly its neigbours also, are currently heavily loaded, that cell cannot be used for adding capacity to another area. Furthermore, the turning of the antenna lobe must be done gradually to avoid the dropping of ongoing calls.
The result of the analysis is sent to a scheduler which is responsible for scheduling the network plans which will be used in the network. Naturally, the analysis is carried out for every cell with detected capacity problems in the network and every section of the time period in question.
4. DATA ANALYSER MODULE (DAM)
The realisation of dynamic network capacity reallocation is implemented in CELLO as Data Analyser Module (DAM). The purpose of DAM is to create a schedule of network plans on the basis of available performance data of that network. Normally the network uses a default network plan, but during peak load situations the network should take another plan into use. The decisions to be made in order to define the schedule are based on the analysis presented in the previous section. The changing of the network configuration is derived from the calculation of a weight parameter called CapacityGain
which is calculated for all MAAs in the network. MAA with the greatest CapacityGain is chosen as the one to be changed in order to provide more capacity to the problem area.
Currently, the performance data in MGIS database is not updated in real-time. Thus the system is not yet suitable for emergency cases. However, it can be used for detecting regular, time dependent changes in the behaviour of the network and for generating adaptive network plan schedules to meet the varying traffic load situations. Furthermore, it can be utilised for demonstrating the Adaptive Coverage System and for enhancing the algorithms used in the analysis.
As stated earlier, the data is not real-time and therefore the time period for which the automatic scheduling is done differs from that of the data used in the analysis. The limits of the time period used in the analysis derive from the time of the OMC data or measurement for which TLMSs are examined. For the analysis to be reasonable, these times must match the time period selected to be scheduled by DAM in some way. The matching can be based on the date, time of day or day of the week.
As described earlier, the analysis has three phases. For the first phase, the detection of cells that have capacity problems, four different factors have been specified.
• Traffic load
• Traffic load (TLMS) • Number of handovers
• A specific message received from network
Traffic load in network cells is obtained from MGIS database, namely the OMC counter which stores the 'Mean Number of Busy TCHs'. Also the number and causes of handovers can be obtained from OMC data and the OMC counter to examine is 'Attempted Internal Handovers Inter Cell Per Cause'. The causes are weighted according to their importance to this analysis. For example, when the cause has been 'directed retry', all the traffic channels in that cell have been occupied and the offered traffic has exceeded the capacity of that cell. In addition, handovers caused by 'distance' indicate that there are some problems in the border areas of that cell and some connections are not automatically transferred to other cells in the area.
The method for the detection of problematic cells is specified for each factor. More precise location for the area where more capacity is needed as well as the detecting of a better network plan are implemented according to the description in the previous section.
The user of the module chooses the time period, granularity i.e. the shortest period of time during which the network plan can be changed and the network plans to be used as well as the factors to take into account in the analysis. The result of the analysis is a schedule for the network plans to be used during the selected time period. An example of the operation of DAM is shown in Figure 2.
5. FUTURE WORK
DAM will be tested in an operational GSM network in a trial which will be conducted in May 2003. Furthermore, some enhancements to the system will be considered in the future. In order to make dynamic capacity reallocation more accurate and adaptive, the data used in the analysis should be as recent as possible. If the data could be analysed in real-time, the changes in the network configuration would follow the temporal and geographical variations of the network load. Thus, the performance of the network could be optimised in all situations.
6. CONCLUSION
The growing requirements of mobile users for network availability, higher bandwith, and reliability, require new, effective ways of network planning to increase capacity. The paper proposes a solution that meets the demands by reallocation of existing resources with minimal investment costs. The proposed solution guaranties localization of capacity problem areas by monitoring such parameters of network performance as blocking rates in traffic or control channels, detection of specific messages (emergency notification), and location-specific handover performance. The obtained information enables a scheduler to choose the most suitable among alternative network plans in order to reallocate capacity to the detected hot-spot area.
The realisation of the dynamic capacity reallocation described in the paper was implemented in the frame of the IST project CELLO, and will be tested during a trial in an operational GSM network. Although the current version of the implementation does not support real-time updates of the system database, it can be used for detecting regular, time dependent changes in the behaviour of the network.
ACKNOWLEDGEMENT
This work has been performed in the framework of the project IST CELLO, which is partly funded by the European Community. The Author wants to thank his colleagues from the CELLO project for cooperation that made the paper possible.
REFERENCES
3GPP, 2002, TS 12.04, “Performance data measurements”, 3GPP
Cello Consortium, Cello Web Page, [Online], Available from: http://www.telecom.ntua.gr/cello/ [20 May 2003]
Eggers, P., Rao, R., Krøyer, B., Kyriazakos, S., Fournogerakis, P., Lähteenmäki, J., Horsmanheimo, S., ACS Requirement
Specifications, CELLO-WP4-CPK-D06, 2001
Jormakka, H., Lähteenmäki, J., Horsmanheimo, S., Pajunen, J., Rissanen, J., 2003, Testing of Location-aided Planning
and Adaptive Coverage Concepts in Operative Mobile Network, to be published in IST Mobile Summit 2003, June
2003, Aveiro, Portugal
Kyriazakos, S., Papaoulakis, N., Nikitopoulos, D., Gkroustiotis, E., Kechagias, C., Karambalis C., Karetsos, G., A
Comprehensive Study and Performance Evaluation of Operational GSM and GPRS Systems under Varying Traffic Conditions, IST Mobile Summit 2002, 2002, Thessaloniki, Greece
