Version du 19 juin 2016 à 22:26

Summary

This article deals with Energy Consumption of Networked Embedded Systems.

Due to the increase of utilization of connected devices, we need to focus on the energy consumption of these connections. Indeed, the Internet of Things allows us to connect our devices to the traditional Internet but the problem is that the protocols we use to do so were made without taking care of energy-efficiency, so these protocols are sometimes not adapted to the software we use.

The first important thing explained by the authors is that of the energy consumption is caused by the radio transceiver.

Authors also explain that there is an energy measurement methodology and there are different categories of measurement and estimation methods according to the device you are using.

Authors then explain the relation between energy consumption and internet protocols like IP, TCP or MAC and then explain the methodology developed in the article. If you use low-power wireless devices, it is difficult to mesaure the energy consumption, so you can only estimate it. They try to explain step by step the process which is based on the use of two energy estimation techniques, by using this method, we want to quantify the differences between simulation-based and real world software-based methods.

In a first phase, authors describe the test Bed description with a MCU, a low power radio chip compliant with the IEEE 802.15.4 standard. There is also an IPv6/IPv4 border router used to connect IP sensor network to a computer. It uses a Serial Line IP (SLIP) to send and receive IP packets with the computer. Then, we need to measure the energy footprint of the Contiki's HTTP embedded web server, we initialize an HTTP request asking for a playload ranging from 1 to 1040 bytes from a HTTP web server.

Then we use a Test Bench to run multiple experiments using two Contiki network configurations. To quantify the motes energy consumption, we must run our tests using two different environments : in the first we deploy the test bed in a real environment, then we use the simulator included in the Contiki network which gives us an ideal experiment environment, avoiding interferences. To characterize and compute the energy consumption for every node on the route when processing a HTTP request, we use the Contiki's built-in energy profiler Powertrace. Every node triggers two Powertrace timestamps, one at the beginning of a request by detecting the TCP opening connection flag and in the second environment, at the end by detecting the TCP closing connection flag.

Then, Authors describe the experimentation phase starting with the simulation of the experiment previously described. The first result exposed is a global equation for the first-hop node when using IPv4 : the energy consumption (in mJ) is : Ev41 = 0,073x + 19. There is also another one when using IPv6 : Ev61 = 0,01x + 20,3.

After that, authors talk about the real world impact on energy consumption, they performed the same experiments with real nodes. The results show that when using IPv4, unlike the simulation based experiments, the energy consumption behavior is more related to the playload size than to the number of transmitted IP packets. In the real world, the energy consumption is higher, furthermore, the standard error deviation is larger following a non constant function curve. The equation of the energy consumption becomes : Ev41 = 0,073x + 21.

Then, authors conclude with the idea that working on the real world rather than on the simulated one requires to handle with an increasing and unpredictable system energy consumption as the transported playload by the WSN nodes grows. This unpredictable and error deviation is caused by the interférences found in real execution conditions.

Finally, we can say that the simulation cannot be a substitute to a real life scenario. WIFI interferences are responsible for this because they impact the radio link quality of IEEE 802.15.4 channels.

Main contribution

This article's main purpose is to explain that even if networked Embedded systems are everywhere, there are still improvements to make. Energy consumption is something becoming more and more important because of wireless communication, we must save as much energy as we can when we use our devices.

The main contribution of this article is the description of the process of calculation of the energy consumption. Authors show that calculate energy consumption is not something easy and reliable because of interferences, this article can also help developers to program a software with all the devices we need and simplify the process.

Applications

The subject of this article is very important in many different fields because energy consumption is something that make a device better than another, so if we can calculate the energy consumption of a device and show that it uses less energy, it will be a better opportunity.

So with an industrial asking, using this process can be a way to prove that a device is better than another. Nowadays, energy consumption is a key-factor of decision when we buy a connected device.

Computering industry can also be interested in that research because it can help developers to make better processors, so it can improve the computer's lifespan.

This article could also be studied to develop an easier way to calcultate the consumption, or to program a software to simplify the process and especially reduce déviations and interferences.

Being able to improve the lifespan of a computer could also be an opportunity to improve other aspects of the computer.

This research can also be helpuf in medical domain especially in medical sensors in human body to signalize sensors in war fields. Being able to control the energy consumption could be a good way to improve that technology and make them smaller.