(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
Oracle 10g R2 and above<br>
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
1 select sid, opname, sofar, totalwork,
2 to_char(last_update_time, 'mm-dd hh:mi')
3 as last_update
4 from v$session_longops
5* where sofar <> totalwork
system@TEST> /
SID OPNAME SOFAR TOTALWORK LAST_UPDATE
---------- ------------------------- ---------- ----------
15 RMAN: aggregate input 108990 124960 02-02 03:37
16 RMAN: aggregate output 65980 0 02-02 03:37
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
system@TEST> SELECT sid, seconds_in_wait AS sec_wait, event
2 FROM v$session_wait
3 WHERE wait_time = 0
4 ORDER BY sid;
SID SEC_WAIT EVENT
---------- ---------- -------------------------------------
1 82491 pmon timer
. . .
12 1006 SQL*Net message from client
<br>
16 0 SQL*Net message from client
<br>
17 1006 imm op
18 1006 sbtbackup
(If applicable)
Author’s Note
Unlike subsequent theories discussed later in this lesson, the limitations of my lab environment did not permit actual testing of the theories in SAME. I am relying on the tests conducted by Oracle and documented in the white paper “Oracle Recovery Manager: Performance Testing at Sun Customer Benchmark Center” for the conclusions described on this page.
Other Text:
(Examples or comments displayed on slide, if any).
Mirror + Stripe = RAID 1 + 0 or “RAID ten”<br>
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
Performance guideline<br>Manageability issue<br>
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
RMAN> # SkillBuilders, 2003
2> # RMAN Tuning
3>
4> # Get elapsed time of full backup (~560Mb) with these specs:
5> # TWO channels
6> # TWO disks
7> # 20 datafiles
8> # FILESPERSET 64 (default)
9> # MAXOPENFILES 16 (default)
10>
11> connect target system/dave@beatles
12> connect catalog rman/rman
13>
14> sql 'begin runstats_pkg.rs_start; end;';
15> run {
16> allocate channel d1 type disk format '/mnt/disk2/%U';
17> allocate channel d2 type disk format '/mnt/disk3/%U';
18> backup full force database; }
19>
20> sql "begin runstats_pkg.rs_stop(P_OP_NAME=>''two_channels''); end;";
21> sql 'commit';
22>
23> exit;
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
sys@TEST> host rman cmdfile=c:\oraclescripts\rman\two_channels.rman
sys@TEST> @display_elapsed_time two_channels
OP_NAME RUN_TIME ELAPSED_TIME ET
-------------------- -------------------- ------------ ----------
two_channels 26-NOV-0312.00.56.1 66701 PM
8364 83.64
sys@TEST> host rman cmdfile=c:\oraclescripts\rman\ten_channels.rman
sys@TEST> @display_elapsed_time ten_channels
OP_NAME RUN_TIME ELAPSED_TIME ET
-------------------- -------------------- ------------ ----------
ten_channels 26-NOV-0312.21.30.0 24133 PM 9906 99.06
Using more channels than output disks increased elapsed time 18%
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
sys@TEST> alter system set backup_tape_io_slaves = true
sys@TEST> scope = spfile;
System altered.
sys@TEST> startup force
ORACLE instance started.
Total System Global Area 236000356 bytes
. . .
Database opened.
sys@TEST> show parameter backup_tape
NAME TYPE VALUE
------------------------------------ ----------- ---------
backup_tape_io_slaves boolean TRUE
sys@TEST> show parameter large_pool
NAME TYPE VALUE
------------------------------------ ----------- ---------
large_pool_size big integer 16777216
Alter parameter
Restart database
Insure large pool allocated
(If applicable)
Other Text:
(Examples or comments displayed on slide, if any).
@v$backup_sync_io
-- TIME TYPE ET EB
-- ----------- --------- ---------- ----------
-- 11/28 11:23 OUTPUT 12900 4362716
@v$backup_async_io
-- sys@TEST> @v$backup_async_io
-- TIME TYPE ET EB
-- ----------- --------- ---------- ----------
-- 11/28 11:46 AGGREGATE 11300 8985827
Async IO doubled “effective” bytes / second. This is GOOD!
However, elapsed time did not drop by ½. Let’s continue tuning.
(If applicable)
