http://www2.computer.org/portal/web/computingnow/focusonwireless

(I know this doesn’t benefit me directly since I already belong to those societies and could probably get access to them through my libraries, but they seem interesting enough and I will likely read at least a few the next chance I get). It’s also helpful for those people who don’t necessarily have normal access to IEEE articles a chance to try them out. Topics include Pervasive Computing, Internet Computing, Security, RFIDs and more. Check it out if you get a chance!

On December 2nd and 4th 2008, Research Day for the Computer Science Department at the University of Guelph is taking place. Research Day is a day where many of the graduate and some undergraduate students present research projects in various stages. For the graduate students the projects consist of initial results and proposals for thesis documents while the undergrads present results from their research projects.

I am scheduled so far to present on December 4th at 12:30 - 12:50 pm. The topic for my talk is not yet finalized however it will likely be something in the area of fair scheduling in wireless mesh networks with multiple gateways. I will report on my progress thus far in my M.Sc. and give a proposal for my remaining work in the next few months. The plan is to have the first draft of my thesis complete by February, the final draft by March and defend by the end of April.

Check back later for another post with the presentation slides and possibly some pictures from research day at Guelph.

Today I completed the listing page for all of the presentations our research group PerWin at the University of Guelph has made in the last year or so. PerWin stands for Pervasive Computing and Wireless Networking Research. The group is directed by my advisor, Dr. Mieso Denko and has several graduate students who are active research members in the group. We have given presentations on topics ranging from Wireless Mesh Networks protocols, Autonomous networks, Fair Scheduling, Load Balancing, RFID, Wireless Security and much more. We have also had special invited talks by several faculty members from various Universities around Ontario (Ryerson, UOIT).

Reynolds Building @ University of Guelph where PerWin meets regularly

Check out the schedule page here for the list of presentations, presenters and topics. There are also listings of many of the websites for the presenters which may be a good resource for people interested in our work. If you are a faculty member interested in giving a talk contact Dr. Mieso Denko at the University of Guelph.

If you are interested in any of the talks I have given, all of the powerpoint presentations are available at my Academic Website at the University of Guelph.

I recently received a request for a guest contribution to the blog from Kelly Kilpatrick so I thought it would be a nice change to have some content from someone other than me. The article is a compilation of some good sites for the latest on wireless technology and news.

5 Blogs and Sites for All Things Wireless

Guest Contribution by Kelly Kilpatrick, October 14th, 2008

As wireless technology continues to advance, it is good to stay informed on the latest technologies, advances, and setbacks related to this convenient technology. However intricate or superficial, these blogs are out there to keep you informed (to one degree or another) regarding the latest and greatest developments in the world of wireless technology.

Engadget: For the latest news and analysis on a wide variety of wireless products and new innovations, engadget has you covered. Read about new hardware and WiMax-compatible products, as well as other developments that are in the works when you visit engadget.

Network World: The contributors at Network World are fully dedicated to finding out everything they can when it comes to the world of wireless. Read comparison reviews, find out the latest news in wireless technology, and learn more about what companies are doing to meet the ever-growing demand for better, faster wireless technologies.

Wireless Week: Wireless Week has several contributors dedicated to keeping the blog up-to-date and fresh. Browse through their substantial archives or look at their latest news and commentary regarding wireless technology. Keeping track of what’s going on in this constantly-evolving segment of computing technology just became a little bit easier with Wireless Week.

WiMax News: To stay current with the latest in this rapidly-growing wireless technology sector, visit WiMax news regularly. Troubleshooting, news, reviews, potential security threats, and much more can be found with a quick search. To stay ahead of the curve, take a look at WiMax News.

MuniWireless: A visit to this particular site will give you a wealth of knowledge related to wireless technology advances and setbacks around the world. Delivering information and articles prolifically, the folks at MuniWireless are certainly in touch with the latest happenings when it comes to wireless technology.

By-line:
This post was contributed by Kelly Kilpatrick, who writes on the subject of dial up internet. She invites your feedback at kellykilpatrick24 at gmail dot com

Introduction

After quite a bit of work on my M.Sc. degree I am beginning to get some initial results that I hope to use in my thesis. My work seems to be headed in the direction of fair scheduling and load balancing in Wireless Mesh Networks (WMN). For more information on why addressing these issues are important and a PowerPoint presentation see my previous post, “Fair Scheduling & Load Balancing in WMN“. Basically, the aim of my work right now is to try to find a fair scheduling such that throughput and fairness are balanced. See the Venn-Diagram I have created below for an idea of where existing solutions fit into this. My solution would ideally fall right in the intersection of the two circles.

Venn-Diagram of Fairness - Throughput Trade-off in WMN Scheduling
© Jason Ernst, University of Guelph

For these results, I have created an event driven simulation in the c++ programming language. The simulation compares the network performance with and without the fair scheduling algorithm enabled. These results are have been submitted to AINA 2009 in a conference paper with Dr. Mieso Denko and are awaiting approval. To gauge performance, two metrics were chosen.

Metrics

• Packet Delivery Ratio:
  is defined as the number of packets that successfully route from source to destination within the network divided by the total number of packets generated. This is an important measure of network performance because a higher packet means that there are less retries on failed packets which can cause congestion in wireless mesh networks.
• Average Delay:
  is defined as the sum of the total delay from source to destination for all the packets. In this simulation, there are two different measures for this: High Delay: which overestimates the delay by including all packets, including packets which only made it partially to the destination and then dropped, and Low Delay: which underestimates the delay by including only packets which successfully arrive at the destination. To calculate the average delay for this simulation, the average between the Low Delay and High Delay was used.

Additionally, the simulation has several parameters. Some remained static for the entire simulation but could be adjusted as part of future work. These parameters include such things are the communication range of the Mesh Routers (MR) and Clients (MC) which was 120m and 60m respectively, and the maximum number of neighbours for a given MR which was 6. Furthermore, there were two variable parameters that were adjusted to yield the results.

Parameters

• Number of Mesh Routers:
  In this simulation the number of MRs varied from 4 to 14 which is a large enough number to cover a significant area in a city while remaining low enough for the scheduling algorithm to compute the schedule in a reasonable time.
• Number of Gateways:
  The number of gateways varies from 1 to the number of MRs in the network. The results presented here are shown with between 1 and 9 GWs.

The network itself is arrange in a square geometry, or in other words an “n x n” arrangement. This is opposed to an “n x m” rectangular arrangement which would likely produce different results. See the Figure provided below for an example of the topology. In the figure, the gateway is in the center of the network, however in the simulation this could interchange positions with any of the nodes. Additionally, as the size grows more MRs join to the edge nodes to expand the coverage of the network. Each GW or MR may have up to six neighbours.

Example of Network Topology for the Simulation
© Jason Ernst, University of Guelph

Results

The following are the results of the simulation. As mentioned above, we will be comparing the performance while varying the number of MRs and GWs in the network. The performance measures are packet delivery ratio and average delay. The simulation was run 10 times and the average was taken for each data point so that consistent results could be presented.

Average Delay vs Number of MRs
© Jason Ernst, University of Guelph

Average Delay vs Number of GWs
© Jason Ernst, University of Guelph

As you can see in the first two figures, there is some penalty in terms of delay for using the fair scheduling approach that we proposed in the paper. However in the next couple of figures you can also see the improvement in packet delivery ratio which is also important.

Packet Delivery Ratio vs Number of MRs
© Jason Ernst, University of Guelph

Packet Delivery Ratio vs Number of GWs
© Jason Ernst, University of Guelph

While these results show greater delay with the fair scheduling approach, it should be noted that there was no attempt at any sort of back-off, retry implementation in either the fair scheduling or the non-fair scheduling (which is used in TCP/IP). If these were implemented there could have been drastically different results. This will be left as future work, however it is expected that the fair scheduling would produced increased performance (in terms of delay) since the avoidance of collisions altogether prevents the back-off and retry from happening as often as it would without the fair scheduling in place.

Future Work

In addition to adding back-off retry mechanisms into the simulation, we would like to incorporate load balancing techniques at the GWs and make use of cross-layered design to further improve the performance of the fair scheduling. Additionally, instead of the c++ programming language, we would like to migrate the simulation into the ns2 or ns3 simulation environment to enable our results to be more comparable to other work in the field.

So I’ve recently come across a bunch of free computers and I’ve decided to try to hook them together to form a beowulf style cluster for running some of my simulations on since I think I can modify the code to be parallel. I have tested all of the computers and have them spread across my basement floor however I seem to have a problem. I need a way to enclose them into some sort of case or storage area. I don’t want to use wood and have been exploring some kind of plexiglass or similar material without much luck. So far the best idea I’ve seen has been the “Microwulf“.

Microwulf Cluster

Anyway, until I start to actually build thing I have decided that I would try to get suggestions through this blog and maybe through some forums on what materials I could use to get these motherboards organized in a reasonable way. (Note: I have extremely limited handyman type skills and a low budget since I’m a student.) Keep in mind that I need a way to not only house the motherboards but also power supplies, a couple of switches and maybe some hard drives. I have 15 working computers so far and do not want to use the standard pc cases because I don’t want a wall of computers in my house. Feel free to offer up any suggestions and check back soon for pictures of the progress!

Update: I have added a project page for the SCrappy Cluster

A couple of days ago while trying to stay awake while my simulations ran for hours, I came across this website: Electric Sheep. It seems like a really cool idea. It is a “distributed screensaver” that makes use of genetic algorithms to progress and improve itself. If the people who are using the screensaver press a button they can rate the screensaver good. This is then used as part of the fitness function in the genetic algorithm allowing the good screensavers to be carried along into the next generations.

Once I get some time I think I’ll try it out. Since I’m on the topic of distributed screensavers… I seem to becoming more interested in distributed systems lately. The one that seems to be fascinating me the most is the DIY beowulf servers. I think this is because I just spent ages waiting for my last round of simulations because of one portion of the simulation that uses clique enumeration / maximal cliques. According to some papers I’ve quickly browsed over this problem is known to be hard to solve so perhaps I could make use of some sort of parallel solution to speed things up.

Not exactly related to computer science, but I thought it would be a helpful post anyways to anyone who might be an Ontario student reading this blog…

I just found out today from Guelph that full-time Ontario College and University students can apply for a textbook / technology grant worth a $100. It may not sound like much but free money is free money and the application takes all of 10 seconds online. That sounds like a pretty sweet hourly rate to me (plus I’m a broke college student myself). Here’s the link if anyone is interested: https://osap.gov.on.ca/eng/eng_osap_main.html. Just click login and look for TTG. According to the FAQ it doesn’t require any proof of financial need or any proof you’ve actually spent the money on textbooks so it’s pretty much no strings attached. For more information about it visit this link: https://osap.gov.on.ca/eng/not_secure/plan_grants_full_sepapp_ttg_12345.htm

Additionally, OGS and NSERC application deadlines are coming up so don’t forget to apply if you are a grad student (or don’t, less competition for me haha)

As an update to my previous post on this subject “Computer Science, A Science?” more debate has taken place on the CompSci.ca forums and blog as a result of a blog post by Tony entitled “The most important programming language”. Many good points were made, for instance: Tony suggested that many people become confused between computer programming and computer science which I agree with.

Perhaps in response to this, Zeroth posted about computer science on his blog with an article entitled “What is Computer Science?” He commented that many people in other disciplines make use of computer science as a way of expressing some of their complex ideas that would otherwise be very difficult to express.

I’m always quite interested in this kind of debate and will continue to follow anything further on the subject. After all the time I’ve been in school and all the profs and other computer type people I’ve encountered I still have to side closest to Zeroth and his post. Computer Science is separate from anything else we have right now. It’s a tool used to help describe complex phenomenon and ideas that could not be otherwise expressed well. It’s not completely math, engineering or an art.

Computer Science is like connection between everything. With computers we can express math through equations and logic within the computer. We can use computers to aid in engineering with CAD and blueprints and models that help with simulation of structures and devices. We can use a computer to create, enhance and distribute music, paintings and video. To me computer science should be trying to do what physics has always aimed for. Since we can use computer science to connect everything we know and understand, we should try to understand everything through the use of computer science. The aim of physics is some magical grand unified theory of everything and I think this theory or “concept of everything” will only be possible from advances in computer science.

Today on a whim I decided to pickup a Linksys WRT54Gv6 wireless router from Staples. It was on sale for $20 off and looked like the last one there. I’ve been looking at the DD-Wrt software for some time since I have an old junky wireless router buried in my closet at home. Unfortunately the old router I have cannot run DD-Wrt so when I saw this one on sale I thought I’d give the WRT54G a try.

Unfortunately, I didn’t realize I bought the crippled version of the good WRT54G since it now has only 2mb of flash memory rather than the 4mb the earlier versions had. Apparently the best version to get now is the WRT54GL because it contains the most flash memory and is meant for Linux. The version I bought has a proprietary O/S installed on it and I was slightly worried I wouldn’t be able to get linux on it, but after following some good instructions at the Scorpiontek website I managed to get it working really easily.

I’ve only have the new firmware on the router for a few hours but I already like it alot more than the stuff that it shipped with. Even with the micro version there are far more options to play around with. The most useful so far for me is the client bridge mode. At my girlfriends house, her Access Point (AP) is located down in the basement and we use the net mostly in her room on the second floor. Until now we had terrible connections up there but in client bridge mode we can make use of the more powerful radio on the AP and plug into it with wires. Seems to be working much more smoothly now than when we had several weaker clients connected in the same room. Perhaps ill try and make some graphs and show if there is an actual difference if I get some time.

Eventually I would like to try to play around with the DD-Wrt source and try to implement some of my wireless mesh ideas in a working device so I’ll see if i can mange with what I’ve got here. Additionally, I have a cool looking LCD panel that an electronics friend gave me so I’ll see if I can get some help hooking that into the router to display some useful information right on the router. I’ll post any progress I make or pictures of mods on the blog here.