Hi everyone,
Following Juha Saarinen's initiative, I decided to myself to some tests
with hdparm at various points in a boxe's "lifecycle". I did these tests
on a Pentium III 733MHz with 512MB RAM and 4 x 30.6GB UDMA/66 7200RPM IBM
hard drives on a RAID 5 array with a 3ware Escalade 6400 controller. I'm
running the latest (well, as of yesterday GMT+0800) CVS of XFS.
Command used for testing: 'hdparm -tT /dev/sda'
I. This was done after the box had been up for quite awhile (overnight)
doing all sorts of normal stuff (installation, basically, as I'm setting
this up as a server). I had about 250MB of my RAM used up, with 66MB in
buffers and 89MB in cache.
---[ results ]---
/dev/sda:
Timing buffer-cache reads: 128 MB in 3.48 seconds = 36.78 MB/sec
Timing buffered disk reads: 64 MB in 3.80 seconds = 16.84 MB/sec
/dev/sda:
Timing buffer-cache reads: 128 MB in 3.45 seconds = 37.10 MB/sec
Timing buffered disk reads: 64 MB in 3.41 seconds = 18.77 MB/sec
Hmm.. suspicious results: probably not enough free memory for a proper test.
/dev/sda:
Timing buffer-cache reads: 128 MB in 3.15 seconds = 40.63 MB/sec
Timing buffered disk reads: 64 MB in 3.21 seconds = 19.94 MB/sec
/dev/sda:
Timing buffer-cache reads: 128 MB in 3.15 seconds = 40.63 MB/sec
Timing buffered disk reads: 64 MB in 3.29 seconds = 19.45 MB/sec
---[ end of results ]---
Please take note of the comment that hdparm gave me on the second run.
II. I rebooted the machine, with all my services (squid, courier, et
al) disabled:
---[ results ]---
/dev/sda:
Timing buffer-cache reads: 128 MB in 0.97 seconds =131.96 MB/sec
Timing buffered disk reads: 64 MB in 1.84 seconds = 34.78 MB/sec
/dev/sda:
Timing buffer-cache reads: 128 MB in 0.96 seconds =133.33 MB/sec
Timing buffered disk reads: 64 MB in 1.77 seconds = 36.16 MB/sec
/dev/sda:
Timing buffer-cache reads: 128 MB in 0.97 seconds =131.96 MB/sec
Timing buffered disk reads: 64 MB in 1.77 seconds = 36.16 MB/sec
---[ end of results ]---
Please take note of the humongous difference!
III. Righter after doing tests in (II), I loaded up a bunch of services
(apache, apache-ssl, courier, postgresql, slapd, samba) and then did the
tests again:
---[ results ]---
/dev/sda:
Timing buffer-cache reads: 128 MB in 0.88 seconds =145.45 MB/sec
Timing buffered disk reads: 64 MB in 1.74 seconds = 36.78 MB/sec
/dev/sda:
Timing buffer-cache reads: 128 MB in 0.85 seconds =150.59 MB/sec
Timing buffered disk reads: 64 MB in 1.75 seconds = 36.57 MB/sec
/dev/sda:
Timing buffer-cache reads: 128 MB in 0.85 seconds =150.59 MB/sec
Timing buffered disk reads: 64 MB in 1.80 seconds = 35.56 MB/sec
---[ end of results ]---
Except for the results of the first pass, the buffered disk reads are
pretty consistent, with the buffer-cache reads going up (I presume because
the cache got populated with hdparm's test data).
Unfortunately I wasn't able to grab top, /proc/slabinfo, and /proc/meminfo
after I did the first (very slow) pass of tests. Next time I get into a
similar situation, I will make it a point to get this information and
share it with the list.
Results of 'free -m' after the tests in (III):
total used free shared buffers cached
Mem: 500 139 360 0 74 18
-/+ buffers/cache: 47 453
Swap: 972 0 972
BTW, for whatever it may be worth, I have quotas and ACLs disabled.
--> Jijo
--
Linux, MS-DOS, and Windows NT ...
... also known as the Good, the Bad, and the Ugly
|