[CentOS] top not showing %CPU

[Bryan J. Smith]

On Thu, 2005-06-02 at 00:42, Warren Harris wrote:
> I'm running with CentOS4 on a 2 cpu machine. I'm running a java process 
> which creates a large number of threads as well as other native 
> processes running under heavy load mixing 200+ audio streams. However, 
> top shows none of these having any %CPU time. All of them read 0.0, 
> except for top itself which occasionally shows up with 0.7, and 
> migration/N with 0.3. All other processes read 0.0. The load average of 
> the machine varies between 1.5 and 4.0.
> Any suggestions on where to look for the problem would be appreciated,

On Thu, 2005-06-02 at 01:08 -0500, Les Mikesell wrote:
> What problem?  Your system should be spending most of its time
> waiting for I/O to complete which doesn't show as CPU activity.

Exactly.

With 200+ audio streams, it's probably a safe bet that I/O is being
slaughtered.  Even "vmstat" won't show those, unless a lot of paging is
going on due to insufficient memory.  "iostat" and similar tools are
what you are interested in.

In reality, I've been meaning to look into what I/O-interconnect
monitoring/stat tools the Linux kernel is capable of.  It's very, very
difficult to gage the amount of I/O Linux can push through systems that
have partial-mesh interconnects (e.g., Opteron).

Some of the newer benchmarks of Linux and Solaris on Opteron aren't
putting a good face on Linux and it's handling of I/O.  It's not really
Linux's fault though.  The PC platform has generally been an extremely
poor platform for I/O for so long, with a single point of CPU-memory-I/O
contention traditionally at the Intel Memory Controller Hub (MCH) aka
"Front Side Bottleneck," so it's really a matter of exposure.

Most RISC/UNIX platforms have it (and lots of tools for monitoring it).
Lintel has traditionally not.


-- 
Bryan J. Smith                                     b.j.smith at ieee.org 
--------------------------------------------------------------------- 
It is mathematically impossible for someone who makes more than you
to be anything but richer than you.  Any tax rate that penalizes them
will also penalize you similarly (to those below you, and then below
them).  Linear algebra, let alone differential calculus or even ele-
mentary concepts of limits, is mutually exclusive with US journalism.
So forget even attempting to explain how tax cuts work.  ;->