Intercell communication
Meanwhile, says Tong, even without the CoS element, hetnets create another backhaul challenge for cellcos - getting all those cell sites to communicate with each other.
"When we talk about hetnets, we're talking about coordinating multiple cell sites to offer higher bandwidth to individual users. That coordination requires some communication between cell sites," Tong explains. "In traditional cell sites, the backhaul is only in two directions - from the cell site to the controller and back, with almost no communication between the sites themselves. But in LTE cell sites are expected to talk to each other via the X2 [application protocol] interface so they decide which one of them can deliver the best connection to a mobile device."
That becomes problematic if the cell sites are linked by traditional point-to-point circuits in a hub-and-spoke architecture, because a base station's traffic has to go through the hub before it can reach the other cell sites it wants to communicate with.
Consequently, says Tong, some operators are looking at using Layer-3 VPN for mobile backhaul rather than a Layer 2 point-to-point circuit.
"If you have two cell sites coordinating together to deliver bandwidth to a particular device, they need to exchange X2 interface traffic frequently, so latency is a key here," he says. "With Layer 3 VPN, cell sites can communicate directly with each other, so in terms of latency it's much better, which is a key point."
SIDEBAR 1: Hetnet: the cost of getting it wrong
Operators planning to deploy heterogeneous networks (hetnets) of small cells to cope with data growth risk investing hundreds of millions of dollars for nothing if they get the business case wrong, according to a new study from Arieso.
The problem, says Arieso CTO Michael Flanagan, is that the hetnet approach creates new problems as operators find themselves faced with multiple offload options, complex subscriber behaviors, myriad charging models and a small minority of extreme users generating the majority of traffic.
"The result is a fiercely complicated business case with some brand new variables," he said in a statement. "The risk of getting it wrong, wasting money and still not solving the problem is significant".
Arieso says its research shows that half of the data traffic generated in a typical mobile network is carried by the busiest 10% of cells, and that one in five of those busy cells have fewer than 200 unique data users, while 2% have fewer than 50 data users.
What that means, says Dr Flanagan, is that cell splitting to provide capacity relief can be unjustified when there are too few users. He adds that even allowing for some high ARPU subscribers, an operator would struggle to justify a macro cell split for less than 200 unique users. Below 50 unique users per cell, Wi-Fi may well be the only cost effective solution.
"In a worst case scenario, where operators try and satisfy the demands of extreme users solely with macro sites, they will waste millions of dollars - $400 million across the industry this year alone," he says.
The $400 million figure is based on data from ABI Research estimating that $2.5 billion in capex is invested by the industry in capacity relief every year, as well as Arieso data that 80% of that investment will go into the areas covered by the 10% of busy cells. (If the remaining 20% is invested in macro sites with less than 200 users, that's $400 million invested in macros with a questionable business case, Arieso says.)
Arieso says that operators need to understand five new variables in their network planning that hetnets bring into play: unique users per cell, how much data they're consuming, device capabilities, spatial distribution of users within the cell, and temporal variations in demand.