Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Mailing Lists -> Oracle-L -> RE: Log writer
Kev,
Pure
shot from hip:
<FONT face=Arial color=#0000ff
size=2>
I'd
say you have more network/app response tuning than log tuning.
I'd
say
you have more disk tuning ( for table/index reads) than log tuning. I
would
DEFINITELY turn the silly ass MTS off.
<FONT face=Arial color=#0000ff
size=2>
I'd
find out how to make sure disk cache is 4 MB and is
writeBACK
not
writeTHRU. ( if either ) <FONT
face=Arial color=#0000ff size=2>I'd have questions about the controller, too,
but.."later".
<FONT face=Arial color=#0000ff
size=2>
(You
didn't tell me the size of the redo log buffer...)
<FONT face=Arial color=#0000ff
size=2>
Of
course, having a two disk wonder (or whatever it is) is the
number
one
thing to work on.
<FONT face=Arial color=#0000ff
size=2>
In any
case, with your hardware, your system should easily be able
to
handle a few users, in its sleep.
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Arial color=#0000ff
size=2>There's about two dozen little tweaks you can do to perk things up a bit,
but
basically, your "rock and a hard place" is going to be a two disk
wonder
on a
single controller...
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Arial color=#0000ff
size=2>hth
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Arial color=#0000ff
size=2>Ross
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Arial color=#0000ff
size=2>
<FONT face="Times New Roman"
size=2>-----Original Message-----From: Kevin Kostyszyn
[mailto:kevin_at_dulcian.com]Sent: Thursday, March 22, 2001 11:31
AMTo: Multiple recipients of list ORACLE-LSubject: RE:
Log writer
<FONT face=Arial color=#0000ff
size=2>Jeez, Seagate Cheeta 18 and 36, 10k RPM
ahhh
UNFORMATTED CAPACITY (GB)
________________22.54 FORMATTED CAPACITY (GB) __________________18.21 AVERAGE SECTORS PER TRACK ________________213 roundeddown ACTUATOR TYPE
____________________________ROTARY VOICECOIL TRACKS
___________________________________167,008 CYLINDERS ________________________________6,962user HEADS
______PHYSICAL______________________24DISCS (3 in)
_____________________________12 MEDIA TYPE _______________________________THINFILM/MR RECORDING METHOD
_________________________PRML 8/9 PR4INTERNAL TRANSFER RATE (mbits/sec)________152 to 231 EXTERNAL TRANSFER RATE (mbyte/sec) _______40
______________________10,025 AVERAGE LATENCY (mSEC) ___________________2.99BUFFER (/optional)
_______________________1MB/4MBRead Look-Ahead, Adaptive,
I believe the only problem is
that they are on one controller. However, this is not a huge database,
only about 3-4 users connected at a time, but lots of DML. The machine
is a Dell Precision 410 with a gig of Ram and dual PIII 450's. I don't
think that it's a huge problem, but it does keep coming up. There are
three log file groups of 32 MB's each and yes the database is also spread
accross these two platters.
Oh and the
results......YOINK!
<FONT face=Arial color=#0000ff
size=2>rdbms ipc
message14980809SQL*Net message from
client9595869pmon
timer1880177virtual circuit
status1877869smon
timer1875449dispatcher
timer1874318SQL*Net break/reset to
client49746control file parallel
write18026db file sequential
read3947log file
sync1456db file scattered
read1017
size=2>EVENT TIME_WAITED---------------------------------------------------------------------------log file parallel
write616Null
event411library cache
pin315refresh controlfile
command292reliable
message273rdbms ipc
reply146control file sequential
read108log file switch
completion58SQL*Net more data to
client 45enqueue27file
identify24
size=2>EVENT TIME_WAITED---------------------------------------------------------------------------log file single
write23latch
free19SQL*Net message to
client17file
open16buffer busy
waits13SQL*Net more data from
client6direct path
read5log file sequential
read0db file parallel
write0instance state
change0direct path
write0
size=2>EVENT TIME_WAITED---------------------------------------------------------------------------LGWR wait for redo
copy0
Disk mfr? rpm? <FONT
size=2>seek/access times? diskperf -y set
already? CPU,I/O, or diskbound system?
log buffer size? log file size? <FONT
size=2>what else is on the "separate" disks? anything? <FONT
size=2>separate disks share a controller? select
event, time_waited from v$system_event order by 2 desc; results?
just a few questions that come to mind....
-----Original Message----- From:
Kevin Kostyszyn [<A
href="mailto:kevin_at_dulcian.com">mailto:kevin_at_dulcian.com]
Sent: Thursday, March 22, 2001 8:50 AM <FONT
size=2>To: Multiple recipients of list ORACLE-L <FONT
size=2>Subject: Log writer
Hi all,
I am testing out
Spotlight on Oracle from Quest Software. It is telling me
that my log writer process is slow. What it looks
like it is saying is that most disks take about 20ms
to right the data and mine is taking on average <FONT
size=2>about 40ms. I don't know why, they are scsi drives and there
isn't that much stress on the syste. Anyway, I
moved the redo logs to seperate disks, but I fear I
may have made an error. Instead of moving each "group" to
seperate hard disks, I moved the members of each group to
seperate hard disks. Anyway, I am wondering if
anyone has advice on this problem.
Sincerely, Kevin Kostyszyn
DBA Dulcian, Inc <FONT
size=2>www.dulcian.com kevin_at_dulcian.com
![]() |
![]() |