This article is more than 1 year old
Understanding the network energy efficiency challenge
Dr Kerry Hinton ticks off seven key energy-saving techs for El Reg
At the end of last week, the GreenTouch telco energy-efficiency consortium told a presumably-glittering event in New York that its five-year project to design more energy-efficient telecommunications has been a success.
In fact, the group said, it reckons that if adopted, its approaches could improve mobile network efficiency by 10,000 times by 2020.
Which is all well and good, but what does it all mean? El Reg spoke to Dr Kerry Hinton, director of the Centre for Energy Efficient Telecommunications at the University of Melbourne, to find out.
The Register's interest was simple: while it's nice to see vendors and academia cooperating, and delivering research, what technologies out of GreenTouch are we likely to see on the street first.
1. Optical interconnect
Hinton, whose specialty is in the core network, started by identifying optical interconnects as an important starting point. Right now, optical links span kilometres or thousands of kilometres, but they're also valuable to cut power over shorter distances.
The growing deployment of optical interconnect “from chip to circuit board to rack” will make a big contribution to telco and data centre energy efficiency (and, after all, work from Intel and IBM bear out the value that big names are putting on that).
2. Router efficiency
“By and large,” Hinton said, “equipment for the core network is built for traffic of a certain type. So a router, for example, might be optimised to get maximum throughput of packets of minimum length.
“That means machines miss the chance to be efficient for different traffic types.”
The opportunity – pursued, for example, in projects like REPTILE (router power monitoring) – is to give up the assumption that “all traffic is short-packet-flat-chat”.
“There are gains to be had by revising the way routers look at packets flowing through the machine,” he said.
The growing penetration of software defined networking (SDN) is important here, he added, because SDN technologies will make the control plane more responsive to different traffic types.
3. Optical transponders
As traffic grows, Hinton said, transponder energy consumption can become dominant in the core network.
One reason is that transponders are designed for worst-case behaviour: “the transponders in the switch or router are designed for fixed-length links – whether it's metres or kilometres.”
It's a problem that becomes more pressing givefn the extra energy used to process traffic at 400 Gbps or 1 Tbps.
If a signal has travelled 1,000 km, “you need a lot of signal processing to repair the signal” – but that's not true if the distance is 1 km, or a couple of hundred metres in a data centre.
Today's transponders react as if “all the signals are equally damaged”, Hinton said. “If you can tune the transponder to do less work, you'll save energy”.
4. Access network
“The biggest issue for most wireline access networks is the home router,” Hinton said – and that's only going to get worse as carriers roll out more sophisticated, and higher-speed, services.
When that happens, the home gateway “has to do more processing and chew up more energy”.
One reason – applicable to the optic networks of the future rather than the limping copper networks of today – is that PON schemes leave it to the home gateway to identify which is “my” traffic on the fibre.
Hence the Bit Interleaved PON (Bi-PON) project: “by redesigning the PON, you reduce the processing speed of the device in the home”, Hinton said.
Instead of the heavy processing deep in the device to throw away other peoples' traffic, Bi-PON makes it easier to discard unwanted traffic – and GreenTouch reckons that can reduce power consumption 30 times.
5. NFV
The other way to make the gateway use less power is to make it less sophisticated. Why have fifty devices in the same street spending five watts each on simple firewall functions (250 Watts) when you probably only need a few watts to run virtual machines in the carrier infrastructure?
“Again, reducing what the CPU is doing [in the gateway] gives you big, big savings.”
6. Small-cell wireless
After giving The Register a caveat that wireless isn't his specialty, Hinton said there's a lot of work happening in reducing cell size.
At the same time, he said, wireless systems that combine MIMO and beam-forming make more efficient use of the radio signal.
A signal meant for Richard can be “beamed to track the user”, which means the base station transmission can be a lot less powerful.
7. Cellular interference
“At the moment, when you get your mobile out of your pocket and you only see one or two bars of reception, that might be because of interference from all the other phones around you”, Hinton said.
The crude workaround today is for the base station to assume it's got to adjust its power upwards – which is bad for energy efficiency.
“If all phones can reduce the power they transmit, you reduce the overall power consumption.”
The University of Melbourne has contributed low signal-to-noise ratio receivers, so that the phones can function in an environment of higher crosstalk.
“That needs a little more processing in the handset, but you can still do that and come out ahead by 10 per cent,” he said.
“And you reduce the power consumption of all the towers, because the phones are less demanding.”
By working right down at the physical layer, “you can redesign the receivers without redesigning the standards,” he added. ®