Database Reference
In-Depth Information
ORACLE version 10.2 trace file. Timings are in microseconds (1/1000000 sec)
Resource Profile
================
Response time: 0.779s; max(tim)-min(tim): 8.902s
Total wait time: 6.012s
----------------------------
Note: 'SQL*Net message from client' waits for more than 0.005s are considered think
time
Wait events and CPU usage:
Duration Pct Count Average Wait Event/CPU Usage/Think Time
-------- ------ ------------ ---------- -----------------------------------
3.834s 491.92% 1160 0.003305s db file sequential read
1.259s 161.54% 1120 0.001124s direct path read
0.350s 44.96% 1191 0.000294s SQL*Net message from client
0.346s 44.40% 10 0.034603s think time
0.090s 11.60% 6227 0.000015s direct path write
0.068s 8.72% 23 0.002954s log file sync
0.060s 7.66% 4500 0.000013s SQL*Net more data from client
0.004s 0.52% 1201 0.000003s SQL*Net message to client
0.000s 0.00% 119 0.000000s total CPU
-5.232s -671.32% unknown
--------- ------- -----------------------------------------------------------
0.779s 100.00% Total response time
Total number of roundtrips (SQL*Net message from/to client): 1201
CPU usage breakdown
------------------------
parse CPU: 0.00s (46 PARSE calls)
exec CPU: 0.00s (47 EXEC calls)
fetch CPU: 0.00s (26 FETCH calls)
Statistics:
-----------
COMMITs (read write): 10 -> transactions/sec 12.830
Note that there are ten instances of think time. This makes sense, since ten iterations were
performed. This means that the profiler has correctly classified the time it takes to access the
JPEG file as think time.
Phase 4—Improvement
It was very obvious that the loading of LOB data needed improvement. From other projects
involving LOBs, I knew that access to LOBs, which are not cached, is quite slow. I also knew that
