RAID 1 Vs RAID 0+1

Question

Hi All,
My database is a part of Oracle 11i application is of version 10.2.0.4 and OS is AIX 5.3.
Our datafile where on RAID 1 configuration (datafile and indexes files on different mount points).
(In our env ,read and write is almost same ,though read is more compare to write)
Due to some io issue (hotspot) and better use of space ,we decided to move our datafiles and indexes file on RAID 0+1 confiuration.
For that we created a volgroup of 4 disk and moved all file to new volgroup.
Suprisingly,the performance started degrading and that hdisk which contain 5 disk started reaching 100% continously.
When I was monitoring the topas command,the 'WAIT' which was intially 0 alway started going to 50 as below
Kernel 2.8 |# | Reads 2593 Rawin 0
User 71.5 |############### | Writes 871 Ttyout 622
Wait 19.5 |### | Forks 7 Igets 0
Idle 17.2 |###########
Can someone guide me,why this happen?or is it bcz my decsion to move the file to RAID 0+1 was bad one or what?
Thanks

Answer 1

>
My OS version in IBM AIX 5.3Have used it in the past - don't have one in front of me,
but bear with me - great system though.
though I dont have any performance report before and after ,I can feel , and users are complaining about the performance.You REALLY* should take a look at NMon - a free tool from IBM itself for
this very purpose - you can even save it for Excel to make nice, lovely,
fluffy graphs for the PHB's.
You should get snapshots (*) every ~15 mins, all day, every day and store them
in some sort of database - look up Nagios with Oracle plugins.
(*) snapshots of iostat, sar, vmstat - whatever's current on AIX these days.
Do something like this
http://www.dba-oracle.com/art_tr_vmstat.htm
" Measure Oracle database server performance with vmstat "
You could also look here
http://www.oracle.com/technology/products/ias/toplink/doc/1013/main/_html/optimiz004.htm
You have neglected to tell us which version of Oracle you are running,
or did I miss it?
You'll have to read and Google and experiment and get a feel for your
own system...
I think WAIT which is going high (which is the new behaviour after the change) is the reason for this.You think*!
That's the problem (and I'm not "picking" on you). You have absolutely
NO* proof that it's your disk configuration - you changed something,
but you don't have a clue how the system was working before the
change, therefore you are in a tricky situation.
Can you change it back to the way it was before, take some metrics
(for at least a month) and then back again?
[prdora@XXXX:/prddb] iostat
System configuration: lcpu=16 drives=15 paths=2 vdisks=0
tty: tin tout avg-cpu: % user % sys % idle % iowait
0.1 50.7 35.7 3.5 56.8 4.0What this is telling me is that your system CPU is idle 56.8% of the time,
not exactly the sign of a system in distress - your data however is worthless,
see below(*)
Check out
http://publib.boulder.ibm.com/infocenter/pseries/v5r3/index.jsp?topic=/com.ibm.aix.cmds/doc/aixcmds3/iostat.htm
(from Google, this you could have done yourself!)
Check out the -D flag to iostat.
Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk1 0.0 0.1 0.0 28678 642459
hdisk12 0.2 36.7 1.0 286506120 10036824
hdisk2 0.0 4.1 0.3 22540800 10820444
hdisk3 0.0 2.8 0.0 22287340 533288
hdisk4 0.0 4.6 0.1 35013920 1981840
hdisk12 is the vritual disk of 4disk in RAID 1+0 configurationThese stats are telling me two things - you have "disk 12"
which is transferring 36.7 Kb per second and is active
0.2% of the time.
This is not* a bottleneck in any man's language - your data
however is worthless - see below(*)
iostats is at taken at an off peak hours(*) I know you have taken your measurements at off peak hours and they are
precisely worthless.
You don't go to the doctor when you're feeling great - why show
us stats from a database that's doing nothing?
When you next post stats, I want to see evidence of a really sick puppy!
Take measurements at peak times - say evey 10 minutes. Do NOT*, whatever
you do, listen to ANYBODY* who tells you that you can't do this as it
will "slow down the server" - you are going to have* to pay a minimum
of overhead to get your data - and it's very small for these metrics.
Try and get an idea of how fast the database is running (for yourself), not
users whining - "the system is slow" is meaningless - maybe they just
want to get home to the girlfriend/boyfriend/whatever...
Run some SQL of your own that takes, say, about a second or so
when the system is quiet - using SQLPlus timing - then run and
capture that SQL and time every 10 minutes - this will eliminate the
need to "feel" anything. Use a statement that always returns the
same result.
Also, lookup the filemon command in AIX - see if one table/tablespace is
very hot!
HTH.
Paul...
When asking database related questions, please give other posters
some clues, like OS (with version), version of Oracle being used and DDL.
Other trivia such as CPU + Disk configuration might also be useful.
The exact text and/or number of error messages is useful (!= "it didn't work!"). Thanks.
Furthermore, as a courtesy to those who spend time analysing and attempting to help,
please do not top post and do try to trim your replies!