Messages in this thread | | | Date | Fri, 7 May 1999 06:48:13 -0700 | From | (Jim Gettys) | Subject | Re: Overscheduling DOES happen with high web server load. |
| |
> From: Andrea Arcangeli <andrea@e-mind.com> > Date: Fri, 7 May 1999 15:22:35 +0200 (CEST) > To: Jim Gettys <jg@pa.dec.com> > Cc: davem@redhat.com, ezolt@perf.zko.dec.com, linux-kernel@vger.rutgers.edu, > greg.tarsa@digital.com > Subject: Re: Overscheduling DOES happen with high web server load. > ----- > On Fri, 7 May 1999, Jim Gettys wrote: > > >I don't specifically remember discussions of TCP RTT estimates, but getting > >better performance data everywhere, including network code, and reducing > >latency in scheduler dependent wakeups of processes (better "interactive > >feel"), as I remember the discussions. Some of the need for this is less > > You get iteractive feeling with reschedule_idle().
Remember, we're talking about different OS's here; not Alpha Linux, which just inherited the higher clock rate. At the time of the MIPS port, Linux wasn't even yet a gleam in Linus' eye, for example.
Also, you may very well NOT be idle; it is feel under load, not interactive feel on an unloaded system.
I think the real reasoning was: 1) didn't take enough CPU time to be measurable 2) had a number of advantages (per previous discussion; better profiling and performance data, better network performance, possibly better latencies on process wakeups, etc.) 3) for the operating systems of the day, it didn't seem to introduce performance problems. If it did, they've long since been detected and fixed. 4) we were scaling up CPU speeds by 10-100-1000 times; while things seldom need to scale linearly, our intuition was that the clock tick rate should scale up at least to some degree.
Again, this is history of 7 and 10 years ago. and some of the reasons then may be moot now.
I think rehashing the history any more is futile; a more serious discussion about whether there are any advantages for Linux to up its HZ rate for the kernel would have merit.
- Jim
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |