|
In this case study, the ROI of an accurate capacity planning process for Universal Mobile Telecommunications System (UMTS)/High Speed Packet Access (HSPA) networks is illustrated. This process is summarized in Figure 7.6, and more details are available in Chapter 4.
After calibrating the system models to be fully aligned with the information in the Operations Support System (OSS), a traffic forecast is generated, which needs to be consistent with the available historical data and include seasonal trends, as well as the impact of planned actions, such as the launch of new services or disruptive marketing initiatives. Based on this process, the forecasted bottlenecks are characterized, i.e. the SON function identifies where and when new capacity expansions are going to be required in order to cope with the forecasted traffic growth at the desired QoS levels. In order to carry out accurate evaluations, initial calibration with current network performance for today’s traffic levels is paramount since otherwise calculations might be biased by unrealistic theoretical assumptions.
In this example, the capacity expansion actions under consideration are: (i) extra carrier additions in those sectors that are expected to be overloaded; and (ii) bandwidth expansions in the Iub interfaces by adding more E1 links at those sites in which the Iub is expected to start causing congestion or excessive latency. Doing this in a fully reactive manner would allow (almost) perfect accuracy, at the expense of degrading QoS due to lack of anticipation, resulting in the initiation of corrective actions when problems are already explicit. However, a p roactive mode in which capacity expansions are planned beforehand is more advisable since it allows more accurate budgeting and prevents the aforementioned QoS degradation.
7.3.3. Modeling SON for Automated Capacity Planning
The SON function under evaluation is the one that executes the calibrated traffic forecast and carries out a qualified assessment of when and where certain capacity expansions are going to be required. Apart from the obvious OPEX savings through automation, this functionality will bring significant CAPEX savings by means of a much more accurate capacity planning process. Therefore, in order to quantify these benefits, it is paramount to characterize the accuracy of automated processes compared to traditional, manual calculations. In this respect, SON processes will be assumed to be ideal, i.e. error-free, whereas manual actions contain implicit errors that are described throughout the rest of this subsection.
Field trials conducted by knowledgeable wireless operators have shown scenarios in which the manual execution of the process under analysis would yield an overestimation rate (o) of up to 42%, whereas the underestimation rate (u) would be around 2%. This means that, in this case, when an engineer needs to manually estimate the amount of additional capacity to be deployed in the network, this figure may be inflated by up to 42% due to forecast/calculation errors. At the same time, while there is idle capacity that has been unnecessarily installed in some parts of the network due to errors in the manual dimensioning process, it is also typical to have certain parts of the network where the installed capacity is clearly insufficient. In field trials, the missing capacity in these under-dimensioned parts of the network has been found to be around 2% of the overall required capacity.
In practice, this over-dimensioning process cannot continue ad infinitum, since the engineers of the operator would obviously notice and correct the situation. For simplicity, the following assumptions are adopted for this analysis of the manual dimensioning process in terms of absorption of capacity surplus and dimensioning error carry-over from year to year:
Only one capacity planning campaign per year is conducted, which is assumed to happen towards the beginning of the year.
All under-dimensioned items are manually detected and corrected the following year by adding the necessary capacity expansions (on top of the additional ones that are required due to the traffic growth that has been experienced). The corrections targeted at fixing the under-dimensioning errors of year i are done during the capacity planning process of year i+1. Since this particular compensation process is reactive, it is assumed to be fully accurate, even when executed manually.
All over-dimensioned items constitute a capacity surplus that will be gradually consumed as the traffic growth in the subsequent years requires extra capacity. No additional expenditure will be carried out until that surplus is exhausted. For the sake of simplicity, full flexibility for HW relocation is assumed, which means that the capacity surplus can be consumed at no extra cost in order to address further expansion needs and manual compensation of under-dimensioned items. Note that in reality this is not the case, since there are HW relocation costs. However, such effect is considered to be out of the scope of this analysis.
Do'stlaringiz bilan baham: |
