Created a simple simulation to reproduce on the emulation network: TCP overshooting a router buffer. Compare using tcptrace.
Output didn’t match. So:
- Fixed up some autoconfiguration stuff in simulation.
- Fixed a bug to do with TCP socket buffer sizes not being the same as emulation network machines in Linux 2.4 due to the memory being “reported” too low.
Now they match fairly well. Haven’t played with Linux 2.6 or FreeBSD yet. Or OpenBSD for that matter.
Also worked on expenses and travel insurance stuff. Unfortunately left my receipts at home, so I need to work on that tonight or tomorrow once I get my folder. Commited a gnumeric spreadsheet for this.
Back in NZ. Again.
Need to reaquaint myself with what I was doing, since I’ve spent some time at the conference concentrating on other things and so far this week I’ve been busy getting myself back into order.
I think I was working on my results chapter, trying to figure out how it would go.
I was going to do more emulation network experiments when I got back into the country I think. Need to figure out exactly what I was going to do here.
Last night after a day in the conference I went back to my hotel room and did a bit of coding to take my mind off things.
So I hacked up my opensolaris stuff and got it linking together after just under 4 hours of work.
Now I need to figure out how to make the code do anything.
Looks like socket creation stuff is in fs/sockfs/socksyscalls.c. Sounds like a good place to start.
Not sure what to do about initialisation just yet. I’ll ignore the problem until it bites me in the * I think.
I’m currently attending the Winter Simulation Conference 2005. I’m sitting in the foyer of the Hilton at the Walt Disney World in Orlando, Florida.
It is now the end of day two, where the highlights were a tutorial session by David Nicol and a neat talk on time warp simulation on 1,033 processors.
I’m a bit bored now. Hence why I’ve got around to blogging. My hotel has no Internet access, so I’m have to do this sort of thing while still at the Hilton.
There are quite a few people about the foyer doing nothing, stuffing arounfd on their laptops. Many wear the WSC badge identifier, but not all. I see a group that seems to be doing some sort of conference using their laptops with someone speaking on loudspeaker. Huh.
Maybe I’ll go in search of dinner later and see if I can find anything worth buying for christmas shopping. Surely I can find something for a laugh.
sim-5 is vaguely interesting. Might be enough to chuck it into the results chapter if I measure it on the emulation network as well.
Changed it around so the ACKs are congested, rather than the data packets.
Looking into fairness measured as they do in “Experimental Evaluation of TCP Protocols for High-Speed Networks, Y. Li, D. Leith and R. N. Shorten, 2005.”
I think I’ve found my TCP flows synchronising a bit at times. Getting some interesting results. Requires further looking at, but may be heading towards a compelling results. Looking at their results, they seem to get synchronisation too. But I’ve noticed my test setup isn’t perfect, when the 2nd flow starts isn’t right in my simulations yet.
Of course, whenever I’ve thought this in the past I’ve been wrong, so I had better not hold my breath.
This is just a technical report. Would be a lot nicer if it were published somewhere good.