Extreme non-uniformity in cellular networks means that usage varies greatly by location, time, subscribers and services. As networks approach theoretical capacity limits, small cells are needed to meet demand. Precise small cell placement is required based on key performance indicators and hotspot identification criteria in order to improve quality of experience.
Today’s radio access networks are massively non-uniform in many dimensions.
Time: enormous growth in demand for connectivity, data rates and volumes. Also more granular changes with levels of demand changing minute by minute.
Subscriber: different cohorts with various usage patterns, tariffs and bandwidth consumption
Location: Huge variations in usage patterns from rural to dense urban. But we find that even within the coverage area of a cell the demand can vary enormously
Service type: voice and data, usage of OTTs, each of which have a different set of requirements on the network to achieve acceptable level of service. Also have different value to the operator in terms of differentiation such as VoLTE.
Let’s consider some of these in more detail.
We know that demand for data is growing fast and that this growth is dependent on service type.
But the demand for data is non-uniform over much shorter timescales
For example, let’s look at the non-uniformities in subscribers and see how much data each user consumes. We can sort them in descending order and calculate their cumulative usage.
Remarkably we see that 90% of the data is consumed by only 10% of the users
Even more remarkably 50% is consumed by just 1%....
This means that if I can add focussed capacity to serve those 1% I have doubled the capacity of my network overall
We took a region that represents many small / mid size countries and broke it into millions of 50m by 50m tiles and then sorted them by consumption by users within the tiles.
Looking spatially like this we see that the traffic is not just confined to a few users but a few places.
No point in just looking at the busy hour – this may change – weekend and weekdays - roamers
So what does all this mean? We know that data demand is growing and will continue to grow in the future and we know that traditional macro networks are limited by the Shannon limit, without more spectrum it is not clear that new techniques will provide the step gains we have been used to.
The good news is, that modern optimization methods are able to target the network performance surgically and with alignment to the operators’ business goals. And when we need to invest in extra capacity, we don’t need this extreme capacity everywhere. If we can find our elusive .35% of the surface and 1% of the customer and serve them, then our existing Macro Network can provide all the requirements for the other 99%.
To be most effective we need to use our subscriber centric data to identify what meets our criteria of being a hotspot and prioritise these. We also want to be able to integrate the small cells with the macro network after placement and ensure that they are optimised.
The calculation and prioritization can be achieved by doing analytics on the subscriber centric data, to combine the different aspects; the mobility, unique users, etc.
This will result in polygons meeting the criteria being identified.
Let’s take a real example. We can study an area using specific criteria.
Here we can see various components of the hotspot definition, for example RF coverage metrics, user counts and degree of mobility.
Each provides its own insight and could on its own be used to identify hotspots
But it’s very difficult visually to determine what are the most relevant hotspots according to all of our criteria in a blended way.
[transition] This is where we can use the power of subscriber centric data in combination with analytics to identify hotspots in a consistent way.
Now let’s look in more detail at this example.
We have identified areas that meet our criteria for being a capacity hotspot.
The next piece of the puzzle is to understand where we have backhaul options.
Now we are fully equipped to answer the question of where it make most sense to place small cells so that we maximise the ROI.
3 example – no hotspot coverage – partial hotspot coverage – full hotspot coverage.
This approach empowers you to surgical place your small cells to maximum effect.