Further evidence has emerged to support the suggestion that operators could take some tactical decisions to meet a great deal of the demand on their network – without necessarily having to move to network-wide upgrades. However, the new research also shows that operators may also have to challenge some common assumptions about small cell deployment if they want to benefit from their
Last month, The Mobile Network carried news from Amdocs’ optimisastion arm that said 20% of locations carry 80% of the traffic on a network.
Today it was rival optimisation vendor JDSU to release some similar looking, but even more extreme, stats. According to JDSU, 1% of all users consumer more than half of all data (Amdocs found that 10% of users are responsible for 80% of demand). But it was the geo analysis that might take you more by surprise.
JDSU carved up each data session it analysed on a “Tier 1 network” into 50mx50m segments* or map tiles, to analyse exactly where users were when data was delivered. What it found is that 50% of all data was being consumed within less than 1% (0.35%) of the geographic area of the network.
Its report states:
“This is a remarkable finding. 50% of the data is consumed in just 0.35% of the 17, 461km2 geographical area covered by this network. 4.47% of the total data is consumed in just the 100 busiest of these 50m x 50m bins. Looked at in reverse, over 90% of the area generates less than 1% of the traffic.”
What is even more surprising is that when JDSU looked into these extreme hotspots, it found that almost half of them were residential locations. It said this counters the often-held notion that residential users automatically default to WiFi usage when at home.
Reasons for this could be LTE’s increased user experience, that the operator offers enhanced or unlimited data allowances and bundles, or that fixed broadband in this particular market is either very poor or expensive.
For JDSU, though, the finding is evidence that challenges the traditional view of small cell planning. It said that targeting small cells, or even Macro expansion, on areas expected to be high traffic, based only on macro cell statistics and local knowledge is likely to be inefficient – as is blanket small cell coverage.
The report stated:
“To effectively plan deployments that ‘personalise‘ the network to meet customer needs requires detailed data about several aspects of data consumption. In particular, networks need to be planned to consider peak data usage as well as total and typical usage, and studying this aspect will undoubtedly reveal additional short lived hotspots. Whether traffic is indoor or outdoor and moving or stationary should also typically be considered when making investment choices.”
* JDSU said the segmentation of connections is vital as a single data connection may last for many hours, with activity spread unevenly between them. Some other techniques are limited to identifying only the start or end location of the connection or context, leading to a “very poor representation” of where the data was consumed. For each of the 7 million geographic bins identified the fraction of the total data consumed in that bin for the 7 days was calculated and the bins sorted by the total amount of data.