Oracle Database Discussions

After upgraded to 11gr2, one job took more than twice as long as before on 10g and 11gr1 with compatibility being 10.2.0.
Same hardware. (See AWR summary below). My analysis points to that Nestloop is doing index range scan for the inner table's index segment,
and then use db file parallel read to read data from the table segment, and for reasons that I don't know, the parallel read is very slow.
AVG wait is more than 300ms. How can I fluence optimier to choose db file sequential read to fetch data block from inner table by tweaking
parameters? Thanks. YD



Begin Snap: 13126 04-Mar-10 04:00:44 60 3.9
End Snap: 13127 04-Mar-10 05:00:01 60 2.8
Elapsed: 59.27 (mins)
DB Time: 916.63 (mins)



Report Summary
Cache Sizes

Begin End
Buffer Cache: 4,112M 4,112M Std Block Size: 8K
Shared Pool Size: 336M 336M Log Buffer: 37,808K


Load Profile

Per Second Per Transaction Per Exec Per Call
DB Time(s): 15.5 13.1 0.01 0.01
DB CPU(s): 3.8 3.2 0.00 0.00
Redo size: 153,976.4 130,664.3
Logical reads: 17,019.5 14,442.7
Block changes: 848.6 720.1
Physical reads: 4,149.0 3,520.9
Physical writes: 16.0 13.6
User calls: 1,544.7 1,310.9
Parses: 386.2 327.7
Hard parses: 0.1 0.1
W/A MB processed: 1.8 1.5
Logons: 0.0 0.0
Executes: 1,110.9 942.7
Rollbacks: 0.2 0.2
Transactions: 1.2


Instance Efficiency Percentages (Target 100%)

Buffer Nowait %: 99.99 Redo NoWait %: 100.00
Buffer Hit %: 75.62 In-memory Sort %: 100.00
Library Hit %: 99.99 Soft Parse %: 99.96
Execute to Parse %: 65.24 Latch Hit %: 99.95
Parse CPU to Parse Elapsd %: 91.15 % Non-Parse CPU: 99.10


Shared Pool Statistics

Begin End
Memory Usage %: 75.23 74.94
% SQL with executions>1: 67.02 67.85
% Memory for SQL w/exec>1: 71.13 72.64



Top 5 Timed Foreground Events


Event Waits Time(s) Avg wait (ms) % DB time Wait Class
db file parallel read 106,008 34,368 324 62.49 User I/O
DB CPU 13,558 24.65
db file sequential read 1,474,891 9,468 6 17.21 User I/O
log file sync 3,751 22 6 0.04 Commit
SQL*Net message to client 4,170,572 18 0 0.03 Network