Sam’s Network Simulation Cradle Blog

I turn 24 today. Hooray for me! The cool thing about this side of the world is, cricket season just started (went to first practice yesterday).

Been getting lots of results for performance chapter. Still more simulations to run! I’ve had to dedicate a decent amount of computing power to this chapter.

Still, I’m not too far away now I think. I suspect it might be another couple of nights before I get all the results I’ve initially planned. This probably means I’ll have been generating results for a whole week, maybe more. That is quite a few CPU hours, given that I do it on a machine in single-user-mode.

Anyway, as those results are almost done, I am starting to put together the rest of the chapter. This requires doing some profiling so I can actually conclude which bits of the code are the bottlenecks.

To this end I’ve set up spectre to run callgrind on a couple of situations. This will allow me to explore what is going on in kcachegrind and maybe even export a callgraph or two if I deem it useful.

I then need to do actual profiling. I’m guessing oprofile will be the tool of choice for that. I suspect I wont get around to that until next week.

I chucked some stuff from my globaliser paper into the perf chapter. I think it looks mostly OK, though the graphs could be fixed up. It is probably enough of a study of the globaliser, I don’t really think I need anything more.

So really, I should be able to have the perf chapter at first draft stage next week. But I’ve got lots of other work to do as well, so it wont be so straight forward.

I need to work on my wns-2 paper, I should ideally get that finished really soon so I get it to Tony. Though I have mostly enough content for that already, it is going to need some hard work to get it into a sensible state, I suspect.

And then there is Intel work, where I have a bit ahead of me still. I think I can stay on top of that alright, things seem to go ahead fairly quickly with that. I’m hacking up my distributed simulation framework to run distributed trace checking stuff. From what I can see, the framework doesn’t need as many changes as I had first thought – I think it should go ahead easily enough.

I’ve written and submitted a paper for MASCOTS’06, put together a chunk of the performance chapter of my PhD, started work on another paper, got NSC partly working in OMNeT++, released NSC 0.2.1, etc.

Been fairly productive, I guess.

There are still some things I would like to get published. So I did a bit of a look around at relevent-ish conferences last night and this morning.

• IEEE Symposium on Computers and Communications (ISCC’06) – Italy – CFP 2005-12-02 – Conf 2006-06-26
• USENIX (USENIX’06) – Boston, MA – Conf 2006-06-01
  • Network Systems Design and Implementation (NSDI’06) – San Jose, California – Conf 2006-05-08
  • Operating Systems Design and Implementation (OSDI’06) – Seattle, WA – CFP 2006-04-24 9:00 p.m. PDT – Conf 2006-11-06
• Winter Simulation Conference (WSC’06) – Monterey, CA – CFP 2006-04-10 – Conf 2006-12-03
• ACM SIGCOMM (SIGCOMM’06) – Pisa, Italy – CFP 2006-02-10 – Conf 2006-09-12
• Workshop on NS-2 (WNS-2′06) – Pisa, Italy – CFP 2006-05-20 – Conf 2006-10-10
• IEEE International Conference on Network Protocols – Santa Barbara, California – CFP 2006-05-09 – Conf 2006-11-12
• Protocols for Fast Long-Distance Networks (PFLDNET’06) – Nara, Japan – CFP 2005-10-14 – Conf 2006-02-02
• IEEE/ACM Measurement and Simulation of Computer and Telecommunication Systems (MASCOTS’06) – Monterey, CA – CFP 2006-04-10 – Conf 2006-09-11
• ACM SIGMETRICS (SIGMETRICS’06) – Frane – Conf 2006-06-26
• IEEE INFOCOM (INFOCOM’06) – Barcelona, Spain – CFP 2005-07-06 – Conf 2006-04-23
• Principles of Advanced and Distributed Simulation (PADS 2006) – Singapore – Conf 2006-05-24

And there is always ACM Computer Communications Review.

There are various measurement conferences I haven’t got here like PAM, IMC.

Journal special issue:

• Special Issue of Computer Networks – Hot Topics in Transport Protocls For Very Fast And Very Long Distance Networks – CFP 2006-05-01 – Publication early 2007

If you ever have 2 of the same codebase, each with modifications – perhaps worked on by two people or something – and you want to interactively go through the changes and merge them, then Meld is the tool for you. Back before I learnt about version control systems, I used to use this method a lot when working with others on a project. Nowadays I have much less use for such a system, but it does come in handy now and again.

For example, I had a large codebase I had been working on and someone else had been working on indepdently. They sent me a tarball of their changes, some of which I had invalidated in my work. I wanted to update their stuff with some of my changes. So rather than stuggle with diff, gvimdiff, patch, etc. I used Meld and it made the job trivial.

Well, once it stopped pausing or crashing when glibc detected invalid pointer usage when comparing directories. I had to set MALLOC_CHECK_=0 to make it go for some reason, maybe something to do with the packing on Ubuntu, who knows. Anyway, once it worked it was great.

Soon to be a new release of NSC. It’s only been 10 months. (Update: yep, 0.2.0 is released).

Finally updated to wordpress 2.0.2, the security release. Should really have done this much earlier. Oh well, it is done now.