Camera ready paper submitted to IEEE GlobeCom 2007

After I worked with my advisor and my co-author Atif Nazir all day yesterday, and they spent the night working on the paper after I departed at 11 pm, and I caught up with them today, we have submitted the camera ready copy of the research paper "A simulation study of GELS for Ethernet over WAN." We found some issues with the simulation results before submission of the camera ready paper and I had to incorporate several fixes to the simulator code with very valuable help and guidance from my advisor and Atif Nazir.

Paper accepted

Our paper on simulation study of GELS has been accepted in IEEE GlobeCom 2007, a premier conference on networking and telecommunications. It is a moment of great joy for us. I am the first author. We owe a lot to Dr. Zartash Uzmi for his tremendous amount of work in the writing of the paper in addition to our own efforts in doing the same. All reviews rated our paper as "well-written."

Executive summary

I've just submitted this executive summary to my advisor. This is related to the summer seminar that every PhD student has to give here at LUMS in the summer. This summarizes the work that we did related to the GELS framework.

Project webpage online

The web page for the LUMS Siemens GMPLS Controlled Ethernet Label Switching project is now live!

Submitted to GlobeCom 2007

It's the 50th anniversary of GlobeCom. It's a premier networking conference. We've submitted a paper about GMPLS Controlled Ethernet Label Switching (GELS) to this conference today. It wouldnt have been possible without the untiring amazing hard work by my advisor Dr. Zartash. I had gone to Sialkot at a crucial stage due to the death of my cousin. I came back the next day, which was one day before the deadline. I had been up all night and driving tired. I slept for only a couple of hours and came for an all night long hard work at Dr. Zartash's office. I slept for three hours at 3 am and then resumed work with him. Atif Nazir was there and he went to sleep later on. Fawaz joined us at around 10 am. By 1 pm, we had completed the paper and submitted it to the conference. My friends later discovered some needed corrections to the paper which were incorporated by my advisor and the paper submitted to the conference again. I'm hoping and praying for the paper to be accepted.

More reading

I'm reading a few more papers these days. One of them is a new traffic engineering approach called COPE, common case optimization with penalty envelop. Another one explores the possibility of doing traffic engineering together with congestion control. Yet another one is about MIRO multipath interdomain routing. Other avenues include virtual network overlays such as VINI. I'll post URLs later on.

Handed over phase II

Yesterday we handed over the results of the phase II simulations of the GELS project to Siemens, Germany. The conference call lasted about 50 minutes. Now we will explore options to further this work.

Paper submitted

This morning, after toiling for weeks on end, we finally submitted a paper to Icc 2007. The paper is title "A simulation study of GELS for Ethernet over WAN." Let's see how the reviews go. It wouldnt be evaluated till January next year.

Corrected

We had very high standard deviations in the convergence time of RSTP in our topologies. Investigating the reasons, we found that links going from Blocking to Forwarding state with a blocked state on the others side were taking two hello time worth in switching state. That doesnt matter to us, because those links dont carry any stations originating traffic (point to point ethernet links) and do not form part of the spanning tree. The spanning tree itself converges upon link failure within a second in all of our failure scenarios. I had to make adjustments to the shell scrip to summarize the results and we are running the simulations yet again.

Bridgesim bug

Running the 50 node topology simulations for single link failures, we came across some situations where the simulator would say, cannot find link between node x and node y, while the simulator itself was showing that link in its log output on the console, right there. Investigating this message, I found that the simulator was checking for NUMBER_OF_PORTS which was #defined to be 40, whereas the port numbers were running all the way to 50. So, I went ahead and changed that in bitmap.h and rebuilt bridgesim. That solved the problem.

Update

So, my advisor is doing this research project with Siemens, which I joined in. The project concerns carrying LAN traffic transparently across a WAN. Not like using VPNs or something. One approach is to use MPLS and carry Ehternet traffic inside MPLS packets. Another approach is to use GMPLS Controlled Ethernet Lable Switching (GELS). Yet another approach is to use Ethernet as is in the core using point to point Ethernet links between the routers.
The problem with the third approach is that it runs STP, which prunes the partially or fully meshed topology, which reduces the available bandwidth and hence less traffic demands can be met. Siemens are interested in GELS (By the way, it is still in the IETF draft stages). But we wanted to quantify this loss in bandwidth incurred by STP, which obviously wouldnt be incurred by GELS since it doesnt prune the topology and will bring other benefits as well.
So, in the first phase of the project, which is already complete, we compared the link utilization, traffic demand acceptance characteristics of STP and GELS and quantified it based on traffic matrices and topologies provided to us by Siemens. We used the TOTEM simulator to do this. We added our own protocol variant of CSPF to the simulator.
The need for the new protocol was to get a common ground for comparison. The problem is that with GELS, CSPF, we are setting up LSPs based on traffic demands, whereas with STP, everything is contention based and there is no concept of LSP. So, we implemented a variant of CSPF that accepts and sets up an LSP if required bandwidth is available and if requested bandwidth for the LSP isnt available, it allocates the highest possible fraction of the requested bandwidth, just as (at some leve), one would get with STP. I'll see if I can upload the code some time.
Now, in the second phase, we are evaluating failure scenarios to see how the two protocols compare in terms of recovery from failures. We are simulating single link failure scenarios for the given topologies on bridgesim. I have modified a C++ program developed by my fellow research student Atif Nazir and developed a shell script to simulate all possible single link failure scenarios with every node in the network selected as root bridge one by one and extracting the convergence time from the simulator's output log file. These results are then put into the gnumeric spereadsheet program and computations are done on the results.
Each single node failure scenario is taking about 50 seconds to run on the Virtual PC environment in which I set up Red Hat Linux 9.0 and there are a total of 82 links in the topology with 50 root nodes, so the entire simulations set will take about 57 hours to complete. Fun stuff, huh? Good for us I wrote the code to automate all this stuff, so that all we need to do is to ensure that the code is doing what we want to do, which I have already done in testing, and to copy and paste the results.

Long absence

I am back after a long absence. It isnt that I wasnt doing anything, but it's just that I was doing too much to be able to report anything here. I was initially reading stuff on running MPLS over OBS for a high speed Internet core and I found some work on it. Presently, I am taking a course "CS 678 Topics in Internet Research" in which we are reading a research paper for every class, which makes for two papers a week. We read the papers on our own, bring a brief summary to the class, where we discuss the paper.
There is also an individual term project for which two other students will be consultants and monitor the progress. For that, we have to come up with a research idea. I have found a few papers on Internet Traffic Statistics and have deposited the proposal to devise a scheme for doing accurate traffic statistics on core routers. We're talking about Tier 2/3 ISPs. I am presently reading a paper about "Inverting sampled traffic."
I made a presentation about it yesterday. The idea is similar to survey sampling. You gather a subset of the data and project information about the whole. I'll give links to documents as we go along.

The basic idea

While I'll refine this technically in more detail later on, the basic idea behind MPLS is that with IP based switching of packets, the routing table lookup is a lengthy process with the longest prefix match algorithm. The routing table must be looked up for every packet destination and the routing table entry that matches to the maximum number of bit positions will determine the next hop neighbor and hence the outgoing interface. While hardware based caching is done to increase the speed of this process by different vendors, with loaded routers, this is going to be sub optimal. So, the idea is that you assign the same tag to all destinations that you switch in a similar manner. Now, even if this tag value is a 32 bit number, the comparison will be orders of magnitude simpler than comparing with all routing tale entries and determining the longest prefix match.
This idea was carried in from the world of deceased ATM, and was implemented in various forms by different vendors such as tag switching by Cisco Systems, and later IETF standardized it to MPLS, which is vastly similar to tag switching, yet different in subtle ways. Also, there exists an open source implementations of MPLS to run on Linux, that I am aware of and is availabe at SourceForge. There might be other implementations and implementations for other OSs as well, but we're not concerned with that.
Note that MPLS does not eliminate the need for a routing protocol. Routing tables still need to be built, but the switching of packets is much faster.

Reading started

I met my advisor Dr. Zartash tonight and discussed my research process with him. He said that he'd be going to Korea for a couple of months or so, and meanwhile, I should read the following two:

• Paper on IP switching by Peter Newman


• MPLS RFC 3031

I am to try to read these before he leaves, (and he doesnt have a schedule for when he will leave) and discuss it with him so that he can tell me where to go next.
I'll keep posting the extracts of what I am reading.