On Tuesday 18 November 2008 16:42, James D. Parra wrote:
Hello,
Trying to find why almost every process that requires IO has near 100% wait times.
Even a simple 'cp' command has enormous wait times and doesn't appear at or near the top of the list when running 'top'. This was occurring when writing to a raid, but have tested writing to the system drive and the same thing occurs.
Cp (and cat) are the extreme of I/O bound programs. They do virtually no computation, just read into a buffer and then write that buffer out. They will always show the lowest CPU utilization and the greatest percentage of I/O wait time. It cannot be otherwise, at least when the file system is hosted by a conventional device (not, say, another file via the loopback or some other non-mass-storage file system).
Running the 'cp command copying a 19GB file from /tmp/file1 /tmp/file2 will take 50 minutes
If I understand you, you're copying the file within a single partition and hence within a single drive, so: 38G = 40802189312 (using CS units; 1K = 1024) 50min = 3000 sec 13,600,730 bytes / sec 26564 sectors / sec That's not impressive, but depending on file system fragmentation and the physical device characteristics, it's not unimaginable. Which file system type is it? What kind of drive? And is the destination really on the same or a separate drive? If it is the same, then it's not really that surprising, since seek locality is almost certainly absent.
The same command on another identical system will take 10 minutes.
Any ideas on what could be causing this?
Top output when running cp;
...
'cp' is not showing up, but it is running;
As I said, low CPU utilization is always the norm for pure I/O applications like cp, cat, dd etc.
root 3915 0.2 0.0 8296 696 pts/0 D+ 15:20 0:02 cp file1 file2
Thank you,
James
Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org