Unsolved
This post is more than 5 years old
9 Legend
•
20.4K Posts
0
1499
October 8th, 2012 20:00
DDBoost and Oracle 10g
We are testing ddboost and Oracle 10g. We have a test database around 40g, installed the latest library RMAN_1.0.1.4-323081_RMAN_linux_64.tar.gz and run a couple of full RMAN backups. First backup took 16 minutes (1 channel), this is a test database with no changes so i figured the second backup will take advantage of boost and complete in no time, that is not the case ..second backup took 12 minutes, third backup completed in 12 minutes again. So what are we doing wrong here, i thought DDBoost for RMAN was supposed to send unique blocks only yet it appears to send the same thing over and over again. Any idea ? Here is our RMAN script
connect target /
sql'alter database backup controlfile to trace';
run {
CONFIGURE DEFAULT DEVICE TYPE TO SBT_TAPE; # default
CONFIGURE DEVICE TYPE SBT_TAPE Backup TYPE to BACKUPSET;
CONFIGURE CHANNEL DEVICE TYPE 'SBT_TAPE'
PARMS 'SBT_LIBRARY=/u01/app/oracle/product/1025/rac/lib/libddobk.so, ENV=(STORAGE_UNIT=ddrmansu01,BACKUP_HOST=dd890.mycompany.com,ORACLE_HOME=/u01/app/oracle/product/1025/rac)';
ALLOCATE CHANNEL c1 TYPE SBT_TAPE
PARMS 'SBT_LIBRARY=/u01/app/oracle/product/1025/rac/lib/libddobk.so, ENV=(STORAGE_UNIT=ddrmansu01,BACKUP_HOST=dd890.mycompany.com,ORACLE_HOME=/u01/app/oracle/product/1025/rac)';
backup database;
backup current controlfile;
backup spfile;
release channel c1;
}
quit
EOF
dynamox
9 Legend
•
20.4K Posts
0
October 9th, 2012 11:00
this is RMAN going directly to DD, no 3rd party backup tools involved.
ble1
2 Intern
•
14.3K Posts
0
October 9th, 2012 11:00
I wish I could say more, but I'm waiting for DD890 to land to our DC where I plan to test the same using NW. I plan to test DFA (client direct) too which I believe is not there in your case (I believe you use TSM, correct?).
ble1
2 Intern
•
14.3K Posts
0
October 9th, 2012 13:00
Ah I see... I will get close to that using DSP and DFA functionality (NW will pick only metadata and be backup orchestrator for all backup operations). I'm not 100% sure, but I believe that with custom library you got you get to see DSP functionality which just pushes part of computing to host side (so in theory you should see a bit more CPU usage on the host), but there is no local cache to have it quick as if it would be client side de-dupe. This is why I don't believe you should see that much of improvements. For a test, have you tried the same without custom library? (using NFS device as disk target and "native" RMAN to disk)
dynamox
9 Legend
•
20.4K Posts
0
October 18th, 2012 10:00
a little bit more testing:
database: Oracle 10.2.5
application: PeopleSoft
db size: ~800G
channels: 6
RMAN full backup to NFS mounted DD890 - ~4 hours
RMAN full backup to DDBoost on DD890 ~ 26 minutes
....amazing, DBA could not believe it so he did a couple db recoveries which worked perfectly (took ~4 hours to restore)
ble1
2 Intern
•
14.3K Posts
0
October 18th, 2012 11:00
Nice... I have variety of Oracle 10 and 11 is all sizes and shapes too. I will first test VTL functionality (this is because we currenly use VTL and I wish to see how it compares). Then I will switch to DD Boost devices and can't wait to test it.