Two-thousand-nine is shaping up to be a watershed year for energy efficiency on data networks. Often overlooked among the eco-conscious, information and communication technologies (ICTs), such as those used in networking products, contribute at least 2% of the world's greenhouse gas emissions.Two key trends are expected to gather speed in coming months: measurement protocols that will enable more accurate energy efficiency measurements across data centers and competing products, and publication of the next draft of IEEE P802.3az, a specification for reducing the energy consumption of networking equipment and communicating state and control information for enabling/disabling energy-efficient modes of operation.

Three proposals have been put forth by Cisco, Juniper, and Japan METI and voting is slated to be completed by next month by the Alliance for Telecommunication Industry Solutions (ATIS) and the International Telecommunication Union (ITU). At issue is how and where -- for example, at the core or at the edge -- to measure the efficiency of networks. Network edge devices typically have loads with less than 3% utilization and high bursts in traffic; enterprise core systems will have a typical load in the region of 15% utilization with fewer, less intense bursts; and WAN core systems are often deployed with typical utilization up to 30%, says Cisco.

Testing must be realistic, reflecting the expected maximum useful bandwidth as well as the amount of power that might typically be available. Taken together, the two measurements can be used to calculate a network's efficiency, says Cisco technical director Hugh Barrass. "It's important that the tests duplicate a traffic profile that is typical for that part of the network," says Barrass. That way, users can purchase and deploy energy-efficient equipment that still takes care of business. "You wouldn't evaluate a vehicle on MPG alone. If you need to carry bricks to a work site, a Prius isn't your best choice."

Meanwhile, the IEEE is making progress on its 802.3az standard, reports Mike Bennett, senior network engineer at Lawrence Berkeley Laboratory and chair of the IEEE's 802.3 Energy Efficient Ethernet Study Group. The study group has been subsumed by the IEEE 802.3az Task Force. "So far, we are on track to complete the standard in the first half of 2010, with draft 1.1 of P802.3az scheduled to come out in a couple of weeks," says Bennett.

P802.3az is the progeny of what was once called Adaptive Link Rate. ALR's aim was to save energy by being able to quickly switch the speed of network links in response to the amount of data that is being transmitted. ALR had its limitations, but Bennett and his fellow engineers feel they are on the way to a solution. He explains:

"The issue with ALR is the amount of time it takes to change from low speed to high speed when the demand for bandwidth increases. Studies showed transition times in the order of 10s of milliseconds, which would be potentially disruptive to VoIP calls, etc. An alternative speed-changing technique, known as subset PHY, was proposed and could change speeds in the order of 10s of microseconds. The term subset comes from the notion that the lower speed would be a subset of the higher speed, e.g., a 10GBASE-T PHY could operate in a 1000BASE-T compatible mode. In November 2007, the concept of Low Power Idle (LPI) was introduced. This essentially takes Ethernet back to the days of 10BASE-T where you send the data as fast as you can then don't send anything. LPI takes this a few steps further in that non-essential parts of the PHY will be power down."

In addition to the energy saved by reducing the PHY (physical layer) power consumption, the LPI protocol can be used to communicate shutdown of other nonessential system components, says Bennett.