In May of 2010, I attended an IEEE conference called the International Communications Conference (ICC 2010) in Capetown, South Africa. This conference is considered the one of the top conferences by the IEEE communications society and is attended by many experts in the communications field. I found one of the keynote speeches by Dr. Steven D. Gray, Head & Vice President Corporate Research, Huawei Technologies to be especially relevant to the current debate in Canada over usage based billing (UBB). The keynote is available free online from the IEEE ComSoc (please note, the keynote I am referencing here starts at about 51 minutes in the presentation).

While the key point about profitability deals with wireless carriers, it may also be relevant to wired network providers. Especially in Canada for the following reasons:

1. The vast, sparse country which must be connected
2. The constant upgrades to equipment required
3. The limited capabilities of backbone and infrastructure networks & increasing speeds of user connections
4. Steady growth in traffic from users

1. The vast, sparse country which must be connected

Compared to other countries in the world, Canada is very sparse. According to a Wikipedia article surveying population density around the world, Canada places 228 in the world. According to this figure (from Gizmodo), Canada ranks 8th in the world for average broadband speeds. Compared with the USA, which ranks 15th (and has a population density almost 10 times higher than Canada), we have less than twice the average speed for roughly the double the cost. Considering our low density and similar geographic size, one would expect we would pay much more.

Map of the Internet [http://www.opte.org/maps/]

2. The constant upgrades to equipment required

The communications industry is constantly evolving as engineers and scientists figure out how to make faster connections or find new ways to communicate (for example, 10 mbps ethernet to 100 mbps to gigabit and beyond), (example 2: Ethernet to fibre optic, or even next generation wireless networks). Every time something new comes out or improvements are made, the companies must put money into making it work with what they already have, deploy new equipment and so forth. Often because the networks are becoming more complex it requires hiring more people to manage them (which may eventually be solved by autonomous networking, but that’s a different story).

Another example of expense to keep up to date in Canada is Bell Canada. Many people with Bell Internet also have wireless access points provided by Bell. Unfortunately, these access points default to (or in some cases only support) WEP encryption. This encryption is not secure at all and can be broken by your average teenager with instruction off youtube. However it could be argued in hindsight this type of expense is the company’s fault.

Furthermore, to increase capacity in the network, it is not as simple as just adding another link between an under-supplied area. Consider a small town where a few people make use of the majority of the connection cause poor performance for the others in the town. This town may be a candidate for increased capacity, but by adding another connection the company does not stand to gain any increase in subscribers and thus may stand to lose money. Now expand this example to larger cities where higher and higher proportions of the population are using more and more of their connection to the point where almost everyone is straining the infrastructure. There are solutions to this type of problem though. Some companies in the US and other places make use of traffic shaping (which while unpopular is a cheaper alternative to the user). Perhaps two types of plans could be put in place, one with traffic shaping and unlimited usage, or one with no traffic shaping and limited usage.

3. The limited capabilities of backbone and infrastructure networks & increasing speeds of user connections

One of the problems with the increased cost to the providers is the increased speed of user connections. Allowing users to have faster connections means they can request more information at once. Imagine many people flushing their toilets at the same time where the pipe at the road isnt big enough to handle all of the water at once. This is what has been allowed to happen. The problem is further compounded because it is not as simple as water simply flowing through pipes. At each junction, decisions must be made on the direction of the flow. If too much traffic arrives at once junction, the time it takes to make a decision is slower than the rate new traffic is appearing and big problems happen.

4. Steady growth in traffic from users

As can be seen in some of the figures in the ICC keynote, traffic from users is growing exponentially. Unfortunately, the service providers’ revenue is usually growing linearly, so there is eventually a point where it is not profitable for the companies to provide Internet service (assuming the cost to provide exponential traffic grows exponentially). This is the most compelling point for usage based-billing. Unless we can somehow find a way to reduce the growth of data, the only way to retain profitable ISPs is to increase the revenue.

Internet growth [Cisco]

5. Conflicting interests between user service and shareholders profits & the disconnect between what you get and what is advertised

This is another important problem with ISPs. Perhaps this is the reason that the infrastructure managed to get into its current state (where it cannot handle everyone making full use of their advertised connections). In order to gain a competitive advantage over competing ISPs, each company often advertises their maximum potential connection speeds. There is never any mention of the network past the point where your house connects to the ISP. Sure if no one else in your neighbourhood is using the connection you might get close to the advertised speed, but as many people know, you often don’t get close. From the shareholder point of view, you want the company to spend as little as possible on infrastructure. Perhaps it may be necessary the enforce a rule that companies must be able to provide the full advertised speed, regardless of how many other people are on the network. Of course, this would make the Internet much more expensive.

6. Potential Solutions instead of UBB

There are many promising technical solutions to some of the problems that are causing the gap between cost and revenue in Internet technologies, however many of them may not make enough of an impact, or are still too premature to cause a difference yet. So in the meantime something must be done.

In the case of providing video services or other multimedia streaming services such as Netflix, LastFM etc. technologies such as multi-casting may be used to deliver common data to a group of users with one packet (in the unicast model, one packet is sent over and over again for every user, even if they are watching the same content).

To reduce the cost of human maintenance and oversight over the networks, applying autonomic computing techniques to networks so that they can self-manage, self-protect etc. may be beneficial.

Exploiting peer-to-peer, caching and other technologies that reduce communications over the large distances on the Internet may help reduce the cost of delivering traffic on the Internet.

Providing tiered Internet service, where users pay for different service levels may be another model altogether that allows companies to remain competitive without usage based billing. Make traffic that causes high strain on the networks (such as streaming video, real-time traffic etc.) more expensive than traffic that is delay tolerant and has lower requirements (web, email etc.)

Perhaps the government itself should take a more active role in providing Internet infrastructure in Canada if the Internet is seen as another piece of infrastructure like roads, bridges and electricity.

Concluding remarks

Of course, there are many assumptions to the case for UBB. The assumption that the cost for service providers is growing exponentially is the biggest and most important. Wired network access doesn’t have the same problems wireless has (the broadcast, limited bandwidth medium, interference etc.) so the comparisons may not actually be valid. The case becomes complicated by the fact that many ISPs are also in the business of providing wireless service, media services (tv, radio, newspaper etc.) that causes conflicts of interest.

The trouble I have with charging the heaviest users is that everyone seems to be trending towards using more and more traffic. In my own experience with large ISPs I feel like I’m always getting less for more money (usage caps have been getting lower and lower, yet the bill keeps going up). The advertised speeds of the networks are going up, but the capacity seems to be falling since we can get more faster, but less overall.

If you would like to see how to install Burg, see my previous tutorial “Ubuntu 10.10 / 10.04 – Grub 2 with Themes” or, a more updated tutorial for installing burg on Ubuntu 11.04 (natty)

I will also make the icons I have created available for anyone who would like to use them. For this tutorial I will use Backtrack Linux, Mobiln and Chromium icons since there is nothing available for them on Burg yet (AFAIK).

Step 1: Edit the burg.cfg file to adjust the “classes”
Notice in the example below I have added the “–class moblin” argument to the entry for Moblin. We will use this later when adding the icons to our templates. You can do something similar for Chromium and Backtrack to make sure the appropriate icons appear for those entries as well.

menuentry "Moblin" --class moblin --class linux --class os --group group_/dev/sda7 {
  insmod ext2
  set root='(hd0,7)'
  search --no-floppy --fs-uuid --set dc4dfece-206f-4d38-b35c-6aac2fb55522
  linux /boot/vmlinuz-2.6.31.5-10.1.moblin2-netbook root=/dev/sda7
}

Step 2: Tell Burg the file names of your icons
Here we will edit the files in /boot/burg/themes/icons. Notice how all the files here are named. We will make our Moblin files: grey_moblin.png, large_moblin.png and small_moblin.png. We will also need to add these names into the “grey”, “hover”, “large” and “small” files.

So, we will just follow the pattern and add entries for moblin. For example, in the grey file, we add the following:

Continue doing this for all the icons you want to make, in all of the types of icon files (large, grey, hover etc).

Step 3: Creating the image
This part is a bit tricky, since burg is a bit picky when it comes to image format. The large images should be 128 x 128 pixels. The small should be 24 x 24 pixels. When you export the png file, make sure you flatten the image if you used layers and then uncheck all of the boxes. You can leave on compression if you like to reduce the file size. The most important thing to note, is greyscale is not supported, so make sure you save back into rgb for your grey files. Also, if you want to emulate how the gray images are smaller, reduce the size to 90×90 and then grow the canvas back to 128×128. (for more details see: http://code.google.com/p/burg/wiki/ThemeCustomization)

Step 4: Testing
If you like you can reboot each time to test out your changes, but the easiest way is with burg-emu. If you run this as root (or with sudo) and give it the “-D” option, you can change themes and view what each looks like with your icons.

The icons…
Here are the icons I made myself. Note: I don’t claim to have any skill in this area, so don’t blame me if they are ugly. I posted this tutorial so that hopefully some people with more skill can create some better ones! If you do, please send me an email and I’ll post them up here, or at least post a link.

  

Download
Here you can get the icons, there is one set for Chromium, one for Moblin and one for Backtract: icons.tar

Sources:
http://code.google.com/p/burg/wiki/ThemeCustomization

Then in GUI select Modem->Connect
Then select “Bell Mobility” when prompted.

Abstract:
In wireless networks, conditions may change rapidly and unpredictably. Often wireless networks are not designed to adapt to these changing conditions and perform poorly when they become congested. The multi-hop broadcast nature of wireless mesh networks amplifies the problem of poor wireless performance. Mixed bias scheduling has previously been applied successfully to wireless mesh networks however, it still suffers from similar problems when conditions change rapidly. In this work we propose an adaptive mixed bias (AMB) algorithm which uses a tabu search approach to change based on delay and dropped packets in the network. The proposed scheduling approach consists of three important algorithms, namely, the tabu search algorithm, move generation, and utility function. The adaptive mixed bias approach is compared against IEEE 802.11 and the non-adaptive mixed bias approach. The performance is evaluated using the packet delivery ratio and average end-to-end delay metrics.

Here are the slides from the talk: BWCCA-NGWMN2010-final (pdf)
and here is the link to the pdf from the conference: Adaptive Mixed Bias Resource Allocation for Wireless Mesh Networks (pfd)

On top of the programming event, we have begun a speaker series at the University of Guelph. The goal here is to get enough people coming out so that we can eventually host an ACM distinguished speaker. Our first speaker was Dr. Dan Ashlock from the department of Mathematics at the University of Guelph. His talk was on video game content creation and was quite interesting. The slides for the talk are available on the ACM guelph website. The turnout was very good for the first talk and we hope it continues in the future. (for anyone interested in attending we also provide snacks!) There are more talks scheduled for throughout the semester from professors, students and industry speakers.

Lastly, we have launched a first version of our website: http://guelph.acm.org. The main purpose of this page is to display upcoming and past events. In the future we also intend on having a vote available for the distinguished speaker we are inviting to Guelph as well as the ability to vote on dates for upcoming events so that we can get the highest attendance possible.

Title: Radio Resource Management in Heterogeneous Wireless Networks

Abstract: Wireless networks are quickly becoming pervasive allowing users to stay connected anytime
anywhere. However, current implementations are not seamless. There are many different radio
access technologies (RATs), for example 802.11 (Wifi), 802.15 (Bluetooth, Zigbee), 802.16 (WiMAX)
and 3g cellular / mobile technologies which are not inter-operable. The aim of heterogeneous
wireless networking is to bridge the technological gap between the existing equipment and software.
When this occurs, the wireless spectrum may be used more efficiently and it is expected users will
have a better experience using the networks. Ideally, the devices should be able to select the best
RAT at a given moment in time based on some criteria, for example cost or capacity. A background
on the architecture of heterogeneous networks will be presented. Problems such as handover, radio
resource management and quality of service (QoS) will be discussed and a direction for further
study will be established.

Update: Here are the slides from the presentation for those who are interested:Research Day 2010 (pdf) – Jason Ernst

While I don’t have a video for you to watch the entire presentation, I do have some pictures, and I’ll post the paper and slides I presented in case any one is interested. If you have any questions feel free to post on here or email me directly and I’ll be happy to answer any of them.

J.B. Ernst and M.K. Denko, “Cross-Layer Mixed Bias Scheduling for Wireless Mesh Networks,” in Proc. IEEE International Conference on Communications (ICC), Cape Town, South Africa, 2010.

Abstract—In this paper we propose a mixed bias approach which makes use of cross layer optimization. The cross-layer parameters are based on conditions in the network from multiple layers and are used to determine resource and time allocation for nodes in the network. Unlike existing proposals, we propose to bias against several parameters such as link quality and queue size in addition to node distance. We also propose a combined mixed bias approach which takes into account multiple parameters together. The scheme is evaluated using simulation experiments. The performance results are reported in this paper.

Download the paper (pdf)
Download the presentation slides (pdf)

Its quite easy to do now in Ubuntu 11.04, really just one step:

sudo add-apt-repository ppa:n-muench/burg 
sudo apt-get update 
sudo apt-get install burg burg-themes

However if you are using 10.04 or 10.10, follow the instructions below:

First, enable the repository for burg by editing your /etc/apt/sources.list file to include the following for Ubuntu 10.10 (Maverick):

deb http://ppa.launchpad.net/bean123ch/burg/ubuntu maverick main 
deb-src http://ppa.launchpad.net/bean123ch/burg/ubuntu maverick main

or use the following for Ubuntu 10.04 (Lucid)

deb http://ppa.launchpad.net/bean123ch/burg/ubuntu lucid main 
deb-src http://ppa.launchpad.net/bean123ch/burg/ubuntu lucid main

Next, (optional) to remove any warnings about gpg signatures, enter the following commands:

gpg --keyserver subkeys.pgp.net --recv 55708F1EE06803C5
gpg --export --armor 55708F1EE06803C5 | sudo apt-key add -

Alternatively, if you find the pgp serer gives you an error, you can try this one:

Then we need to update apt and install burg:

sudo apt-get update
sudo apt-get install burg-pc burg burg-themes

This will prompt you to select several options along the way, so far I’ve just selected the default options since they seem to be detected from your existing grub install. The important one is to select the correct disk (Note: I’ve only tested in a non-raid system, so I don’t know how it will behave with this setup).

On the next restart, we should see a graphical grub menu, something like this:

Ubuntu 10.04 with Graphical Grub Template

By default you will probably get the ubuntu template, and you can change the template by pressing ‘t’ during the grub screen. It should also remember which one was last selected.

Note: I have experienced some problems with a fresh install of ubuntu 10.04 and SLI video cards. Burg’s graphics mode seems to confuse ubuntu and it gets stuck on boot. If this happens, you can boot in recovery mode and install the restricted nvidia driver. Then on next boot everything should be fine.

Note 2: Occasionally, when Ubuntu updates, grub may install over your burg installation. In order to get burg back, you need to issue the following command:

where /dev/sda is the partition you want to install burg onto.

Note 3: You might find you have some extra entries that you want to remove, for example the recovery entries or whatever else. You can either edit the burg.cfg file directly located at /boot/burg/burg.cfg and or you can edit the files in /etc/default/burg and /etc/burg.d/. If you use the second choice, you will need to run:

in order to regenerate the new burg.cfg file.

Sources:
https://help.ubuntu.com/community/Burg
http://ubuntuforums.org/showthread.php?p=9231199