Fueling Wireless Networks

Jerry Zhou
December 9, 2011

Submitted as coursework for PH240, Stanford University, Fall 2011

Bit by Bit Consumption

With the implementation of IEEE's 802.11g wireless networking standard, the transmission and reception rates of everyday wi-fi devices have began to approach the order of the theoretical limits proven by the Shannon-Hartley theorem. As a consequence, it is much more practical at this time to accurately discuss the long-term energy needs of a global wireless network.

All wireless data is sent in packets. While packet sizes vary, the basic unit of a data packet is always the bit. With this in mind, we can model the energy usage of a wireless network based on the total number of bits processed. One proposed method is to describe the energy consumed by a network with the equation Energy = k × bits processed + fixed cost. [1] Where the constant coefficient k is based upon the physical hardware and the fixed cost is associated with the transmission rate. At the expense of discounting additional energy consumption due to bit errors, the linear model of energy use has been experimentally confirmed to be a good fit, showing a .95 or higher coefficient of correlation in many tests with commonly available networking interfaces. [2]

One such experiment was carried out to determine the k and fixed cost values for Lucent IEEE 802.11, 2 Mbps WaveLan network card, a generic and widely used network card installation. The experimenters made efforts to circumvent the issues of channel noise, avoiding problematic bit errors. Results were tabulated for both broadcast and receive modes. The energy cost for broadcast mode was 1.9 μW-sec / byte + 454 μW-sec. The energy cost for receiving was much less, at .5 μW-sec/byte + 356 μW-sec. By comparison, the idle power usage for the same network card was measured at 66 mW. [3,4] From these figures, one can see that the total energy usage is dominated by the fixed costs when receiving and transmitting at current data rates.

Creating an Efficient Network

While industry and the IEEE continue to attempt at developing higher transmission rates, the impact on the per-bit energy consumption has stayed relatively small. In order to drive down the energy usage of wireless devices, efforts must be made to reduce the fixed cost of transfer.

Specifically, this fixed cost comprises of constantly sensing incoming transmission, accessing synchronization messages, and intermittent control transfer between base station and receiving device.

One approach to reducing this fixed cost is to improve hardware utilization. However, each percent reduction in energy cost from hardware change alone demands a significant increase in manufacturing costs. [4] This extra cost in addition to the perceived miniscule energy requirement for any single device has prevented an active movement for an energy conscious materials and hardware.

However, aside from any hardware changes, researchers have recently discovered additional methods to reduce the fixed cost associated with data transfer. One such method is the usage of an "ad hoc" network in place of a "bss", or base station subsystem network. An ad hoc network is one in which every receiver node also acts as a transmitter or forwarding point for other nodes. This is contrasted with the preeminent networking schemes which involve a centralized base station that services all of the surrounding receivers. Currently ad hoc infrastructures suffer from the additional complexity and errors involved in the multi-hop routing of information from one receiver to the next. In addition it also carries with it a higher idle energy cost for receiver devices compared to receivers operating in bss mode.

As routing protocols for ad hoc networks become further developed, these networks stand to become a much more energy efficient solution than the bss network. The direct energy use benefits of the ad hoc network include a much lower detection threshold, reducing the fixed cost associated with sensing transmission. Furthermore, no control transfer has to be done between the base station and the receiver, since each receiver is already authorized to be a transmitting node as well. In experiments performed with the 2Mbps WaveLan card operating in ad hoc mode, the idle energy cost increased to 184 mW, but the fixed cost was reduced to 56 μW-sec for receiving, and 266 μw-sec for transmitting. [3] As data rates climb to the 802.11g standard maximum of 54Mbps, the energy savings of an ad hoc network begin to be quite significant.

Conclusion

By introducing additional complexity into the network scheme, there is the potential to create a much more efficient way to transmit data. The issues surrounding a functional ad hoc wireless network is still an obstacle to realizing a more energy-efficient network, but classes of protocols are already being laid out that resemble an intermediate between ad hoc and bss networks, assigning special nodes that act as temporary base stations and rout information a few hops instead of direct transmission. As more and more applications begin to turn wireless, a reduction in the fixed cost associated with data transfer allow these wireless networks to remain a manageable source of energy consumption.

© Jerry Zhou. The author grants permission to copy, distribute and display this work in unaltered form, with attribution to the author, for noncommercial purposes only. All other rights, including commercial rights, are reserved to the author.

References

[1] J.-H. Chang and L. Tassiulas, Energy Conserving Routing in Wireless Ad-hoc Networks," Proc. INFOCOM 2000 (IEEE, 2000), p. 22.

[2] J.-C. Chen et al., "A Comparison of MAC Protocols for Wireless Local Networks Based on Battery Power Consumption," Proc. INFOCOM 1998, (IEEE, 1998), p. 150.

[3] L. M. Feeney and M. Nilsson, Investigating the Energy Consumption of a Wireless Network Interface," Proc. INFOCOM 2001, (IEEE, 2001), p. 1548.

[4] "WaveLAN 802.11b Chip Set for Standard Form Factors," Agere Systems, PB03-025WFC. December 2002.