The Blog

Since this took place, I have sort of decided real time was a good option, and so ego and I had yet another one of the crazy conversations…

(11:35:18 AM) Dhaval: anyway, the headache got diagnosed as a tension headache by the doc
(11:35:31 AM) Dhaval: so dipankar gave me the “women trouble” talk
(11:35:39 AM) Gautham: ah, okay
(11:35:42 AM) Dhaval: dude. the most embarrasing few mins i have had in a few days
(11:35:45 AM) Gautham: tension
(11:35:52 AM) Gautham: what happened ?
(11:35:55 AM) Gautham: woman trouble ?
(11:36:08 AM) Dhaval: there should be woman to cause woman trouble
(11:36:18 AM) Gautham: well, isn’t there one ?
(11:36:18 AM) Dhaval: (though there is one now, and she ain’t causing any trouble yet!)
(11:36:25 AM) Gautham: i thought you had a multi-queue scheduler
(11:36:39 AM) Gautham: well, trouble comes in various forms
(11:36:41 AM) Dhaval: did you not see my last status?
(11:37:03 AM) Gautham: nope
(11:37:19 AM) Dhaval: it was “scheduling real time now”
(11:37:32 AM) Dhaval was doing fair group scheduling earlier
(11:37:42 AM) Gautham: I can understand
(11:37:47 AM) Gautham: when you have realtime tasks
(11:37:58 AM) Gautham: the other tasks, which are performance oriented and long running tend to suffer
(11:38:01 AM) Gautham: have experienced this
(11:38:07 AM) Gautham: and it wasn’t a great experience
(11:38:22 AM) Dhaval: oh man
(11:38:23 AM) Gautham: so, i tend to run them on different boxes thes days!
(11:38:27 AM) Dhaval: the women troible talk from you as well?!
(11:38:47 AM) Gautham: Just passing on the experience so that human species evolves without repeating mistakes
(11:38:51 AM) Dhaval: yeah. i have started limiting myself to just 3-4 these days
(11:39:26 AM) Gautham: In my case it was worse
(11:39:38 AM) Gautham: the realtime jobs were long running as well as latency sensitive!
(11:39:44 AM) Gautham: which means, they’re going to hog the machine
(11:39:46 AM) Gautham: complete
(11:39:47 AM) Gautham: ly
(11:39:50 AM) Gautham: no time for other interrupts
(11:39:50 AM) Dhaval: dude. see, that is why you do the fair scheduling
(11:39:56 AM) Gautham: well, not for rt!
(11:40:00 AM) Gautham: you simply can’t
(11:40:05 AM) Gautham: i was running a -rt kernel
(11:40:12 AM) Dhaval: i know. you cannot have too many rt processes running
(11:40:18 AM) Dhaval: remember, you have only 1 core :/
(11:40:23 AM) Gautham: i had one -rt long running process
(11:40:46 AM) Dhaval: you need one whcih can take a few deadline misses once in a while
(11:40:47 AM) Dhaval:
(11:40:49 AM) Dhaval: a robust one
(11:40:51 AM) Dhaval: get married
(11:40:54 AM) Dhaval: it will allow that to happen
(11:40:54 AM) Dhaval:
(11:40:55 AM) Gautham: heh!
(11:40:59 AM) Gautham: yeah, true!
(11:41:03 AM) Gautham: which i didn’t want to do
(11:41:08 AM) Gautham: hence changed the kernel back to fair sched
(11:41:10 AM) Dhaval: so then you better get used to the strict latency
(11:41:10 AM) Gautham: no rt
(11:41:11 AM) Dhaval: heh
(11:41:15 AM) Gautham: and interrtups are not threaded!
(11:41:18 AM) Dhaval: poor tglx won’t like that one
(11:41:21 AM) Dhaval: one bit!
(11:41:24 AM) Gautham: well, ingo will!
(11:41:31 AM) Dhaval: i think ingo is preempt-rt dude
(11:41:41 AM) Gautham: okay vatsa will!
(11:41:44 AM) Dhaval: heh
(11:41:54 AM) Dhaval: i am no sure vatsa is interested any longer
(11:41:58 AM) Dhaval: he prefers co-op scheduling
(11:42:04 AM) Dhaval: which would happen after marriage and kids i am sure :/
(11:42:35 AM) Gautham: heh!!!
(11:42:51 AM) Gautham: well, one has to have multiple cores to be able to do that!
(11:42:54 AM) Gautham: it’s advanced stuff
(11:44:19 AM) Dhaval: and you believe vatsa is not advanced enough?
(11:45:27 AM) Gautham: he is! I am not

And after all this analysis, I wonder, why is it that we find rt too hard :/