Hi there Patrick, Sorry to take so long to reply. As you correctly point out debugging and probing slow things down. Unfortunately with any debugging tool there are side effects which can, at times, render things useless - a bit like Schrödinger's cat. So use this for debugging when you need, or get a faster machine to keep up with the debugging. My tendency for high speed bugs is to just put in printk(KERN_CRIT "blah") into the code when I'm working on a bug and only spit out the information I want and no other debugging. Another tool that can be quite useful is to use tcpdump and then analyse with WireShark (formerly Ethereal). I prefer to directly capture using tcpdump first as less overhead. Also in my testing I shut down all unnecessary services such as X, apache, exim etc as these can have quite an impact on performance. The biggest impact is if you're trying to use dccp_probe from a graphical window as opposed to a console. Anyway I'm rambling a bit and hope this is of some use. Ian On 6/6/07, Patrick Andrieux <patrick.andrieux@xxxxxxxxx> wrote:
Hi again, Here is a description of another point I'd like to report. It's obvious if you think on it, but it's not if you don't. I ran some tests the last couple of days, and I was surprise by the bitrate iperf gave me : at about 50kbits/sec... and I didn't understand why since last week. The problem was the debug messages printing. It takes lots of time to write in a file, and I enabled debug messages options. So, I can let you imagine my logs... 3Go at the end of the day... I disabled these options (in /etc/modprobe.d/networks)... the bitrate given by iperf is now about 2.3Mbits/sec. And I noticed the same "problem" using dccp_probe module only. It slows down the traffic at about 1.7Mbits/sec. It prints out less information so it makes sense to have a better birate than with debug messages enabled. So the point is all these values are wrong since their printing affect rate, RTT, loss, .... calculation. rgds, Patrick. - To unsubscribe from this list: send the line "unsubscribe dccp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
-- Web: http://wand.net.nz/~iam4/ Blog: http://iansblog.jandi.co.nz WAND Network Research Group - To unsubscribe from this list: send the line "unsubscribe dccp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
