> iostat -cxnz 1 201 gives the following indication:
>
> cpu
> us sy wt id
> 22 78 0 0
> extended device statistics
> r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
> 0.0 43.0 0.0 159.5 0.0 0.1 0.1 2.3 0 10 c1d0
> 0.0 43.0 0.0 159.5 0.0 0.1 0.1 2.0 0 8 c2d0
What is your exact system setup, and what test are you running?
> So, I guess that the high cpu usage combined with low
> disk utilization (%b) could indicate something,
> right? Am I looking at kernel driver issues or what?
> I'm using an Asus M2NPV-VM motherboard with an Nforce
> chipset.
Well, I have the same Asus M2NPV-VM mainboard, and
a zfs mirrored poll with two S-ATA drives:
% zpool status files
pool: files
state: ONLINE
scrub: none requested
config:
NAME STATE READ WRITE CKSUM
files ONLINE 0 0 0
mirror ONLINE 0 0 0
c4d0s6 ONLINE 0 0 0
c5d0s6 ONLINE 0 0 0
errors: No known data errors
Creating a 4.8GB file on that zpool with...
% time mkfile 4800m foobar
0.01u 2.51s 0:04.46 56.5%
... shows iostat rates that are similar with the numbers you've got:
us sy wt id
1 54 0 44
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
0.0 103.1 0.0 365.8 0.1 0.1 0.8 1.2 1 11 c4d0
0.0 100.1 0.0 365.8 0.1 0.2 1.0 1.5 2 13 c5d0
cpu
us sy wt id
0 64 0 35
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
0.0 99.1 0.0 378.2 0.1 0.1 0.6 1.2 1 11 c4d0
0.0 98.1 0.0 377.2 0.1 0.1 0.6 1.3 1 12 c5d0
The "mistake" with this "benchmark" is that zfs compresses away
all the null bytes; not that the file consumes just a single disk block,
although it has a file size of 4.8gbytes:
% du -h foobar
0K foobar
% du foobar
1 foobar
% ls -l foobar
-rw------- 1 jk usr 5033164800 Apr 3 11:16 foobar
All that zfs has to write are a few file metadata updates (modification
time updates, file size updates), and that results in a few hundred
kilobytes written to disk per seconds.
Note also, that when you compute the number of bytes written
per second by mkfile you get a transfer rate at the filesytem
level of > 1 gbyte/sec!
% bc
scale=3
5033164800/4.46
1128512286.995
If you want to measure / verify SATA performace, run the
"iostat -cxnz 1 201" command in a window, and at the same
time read from the raw disk device, using something like
dd if=/dev/rdsk/c1d0p0 of=/dev/null bs=32k
When I run such a test on the Asus M2NPV-VM, I see
cpu
us sy wt id
1 12 0 87
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
2407.0 0.0 77024.6 0.0 0.0 0.8 0.0 0.3 4 84 c4d0
cpu
us sy wt id
0 11 0 88
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
2427.0 0.0 77664.3 0.0 0.0 0.8 0.0 0.3 4 85 c4d0
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
