Re: iostat patch: Add dillon's systat changes to iostat
Matthew Dillon schrieb:
> * Provide options for various modes, with the default being the old
> (KB/t, tps, MB/s) tuple, so the output can remain compact.
> I recommend one that does (tps, MBr/s, MBw/s).
Yeah, thought about that, too. Will look into it.
>
> * Maybe we can remove the tty statistics and add aggregate or
> primary interface network statistics. Nobody cares about tty
> statistics any more but it would be kinda cool to see network
> stats there mixed with the disk stats.
Can I get that from devstat also? Well, could just look at systat
-ifstat, right? How to determine the "primary" interface?
If you don't want to see tty (which I fully understand :-) ) you could
just use iostat -dC.
> I am considering removing the tps/r, tps/w from systat -vm 1 and
> just having tps, but keeping MBs/r and MBs/w. I think people will
> get more out of an aggregate tps stat. Then for systat -vm 1
> we would have: (KB/t, tps, MBs/r, MBs/w) (I should rename those to
> MBr/s and MBw/s I guess too).
Yep, Megabyte-seconds divided by writes (and reads) doesn't make a
horrible lot of sense :-)
Jan
討論串 (同標題文章)
完整討論串 (本文為第 4 之 5 篇)：