Subject: Re: [linux-audio-dev] info point on linux hdr
From: Jussi Laako (jussi_AT_jlaako.pp.fi)
Date: Sun Apr 23 2000 - 12:09:05 EEST
Benno Senoner wrote:
>
> SCHED_FIFO + lowlatency patches is immune from disk I/O .. see my pages.
...but nothing to do if that app goes nuts. SCHED_FIFO scheme doesn't fit
all applications and SCHED_RR is too stupid. SRT-Linux looks promising,
haven't tested it yet though.
My app receives multiple audio streams over network, applies some filters
and interleaves data to single output stream played through single sound
card. That's not problem, but gets more hairy, because if one stream stalls
it may not stall audio output, other streams must continue to play. Latency
must also be low to keep phase shift between streams low. (You can imagine
how it sounds if there is 16384 sample "phase shift" between streams because
of buffering!)
> anyway as soon your machine is in the TRASHING status, lowlatency or not,
> the whole box gets unusable.
I think this is a security risk, kind of DOS attack... And easy to do with
user privilege process.
Kernel should measure time (in %) it spends swapping any processes pages, if
that starts to reach 100% of CPU time it should just kill the process or at
least limit the time to say 90%.
> As for IO hogs, the problem is that they execute in kernel most time,
> therefore regular processes suffer big.
Yes, I'd like to tell kernel that it can spend 75% of it's time on sound
card, 20% on network IO and 10% on disk IO.
Also that process x can spend maximum 50% of CPU time and 25% of total IO
bandwidth and that it's scheduling latency requirement is 25 msecs...
Umm, now I'm asking something that only mainframes can offer. ;-)
Anyway, I hope SRT-Linux can do what it promises.
- Jussi Laako
-- PGP key fingerprint: 161D 6FED 6A92 39E2 EB5B 39DD A4DE 63EB C216 1E4B Available at: ldap://certserver.pgp.com, http://keys.pgp.com:11371
This archive was generated by hypermail 2b28 : Sun Apr 23 2000 - 12:41:12 EEST