Focus area: Supply and control

Hurry up and WAIT

Project URL:

Solution/product description

The project combines aspects of measurement and "supply and control".

The project has two parts: first, we will be setting up a lab that allows both local and remote participants to experiment with DOCSIS systems. Currently, building an HFC system requires resources, both financial and in technical expertise, far above what's within reach for most academic labs, shutting out almost all contributors. (Indeed, there seem to be very few papers on the general topic of end-to-end energy consumption written by academic research institutions. Using the experience with NSF GENI-style networks, authorized experimenters should be able to get temporary remote access to the HFC infrastructure, update code and run measurements. As cloud-based CCAP become more common, slicing such systems may become feasible as well.

The second part explores the notion that residential usage is both bursty and (somewhat) predictable at various time scales, allowing cable modems to enter various depths of sleep mode. We divide the operation of a home network into three modes: (1) high-intensity, "full on"; (2) low-intensity notifications and background; (3) idle. High-intensity mode occurs when downloading video chunks, for example, or large files (e.g., system upgrades). Low-intensity mode sees brief bursts of packets, e.g., IoT notifications, VoIP signaling or web-based push applications. For high-intensity services, they are often extremely bursty, with, say, a one-tenth second gigabit/second download followed by nearly ten seconds of silence. For low-intensity modes, much of the communication is not latency sensitive, i.e., can tolerate delays measured in seconds or even minutes.

The achievable energy savings depends on whether reducing the PHY capacity to one 6 MHz channel (or, in the OFDM future, other channel widths) and sleep modes can significantly reduce energy consumption. Thus, the first task of the project is to measure the relative impact of adjusting the available capacity and of putting systems or parts of systems (e.g., parts of the remote PHY) into short-term sleep modes.

Longer term, we envision a cooperative scheduled system where in-home systems are provided wake-up times to bundle their transmissions. Since all such systems now have accurate time information via NTP, they can schedule their delay-tolerant transmissions and inform their controlling servers to do the same. For existing systems, this won't be possible, but since many background systems are highly predictable, the cable modem should be able to learn their traffic patterns and schedule wake-ups accordingly. Also, given TCP SYN retransmissions, many server-to-home transmissions will work even if unpredictable, even if not on the first try.

How does your project address / improve the focus area (Monitoring and Measurement, Demand Response, Supply and Control)?

The project integrates measurement, protocol design for schedulable transmissions, and machine learning to estimate and modify transmission schedules.

What was your inspiration for your project?

Mobile systems, particularly in paging mode, also rely on periodic wake-ups.

What is the single most exciting thing about your project?

Having shareable infrastructure will make it possible for many more teams to contribute to this area, with the near-insurmountable burden of setting up their own CCAP and cable modem system.

Why did you decide to take part in the Adaptive Power Challenge?

I believe that as other parts of the home, such as lighting, become more energy-efficient, communication services will likely make up an increasing fraction of the residential energy budget, as they are on 24x7. While the individual energy consumption per user is probably modest (10 W per modem, plus the Wi-Fi AP and the head-end), this is multiplied by roughly 88 million households in the US. According to the NRDC 2013 study, modem usage for about half the residential Internet households corresponds to 2.0 TWh per year, more than for TVs.

Also, energy efficiency will make it far easier to sustain broadband operations via battery backup during power outages, without overloading the cellular network.