Diffrence between cpu and elapse time in tkprof

Question

Hi All
i found huge diffrence between cpu and elapsed time in tkprof. can you please advice me on this issue.
>call count cpu elapsed disk query current rows
==================================================
Parse 1 0.12 1.36 2 11 0 0
Execute 1 14.30 720.20 46548 190520 205 100
Fetch 0 0.00 0.00 0 0 0 0
======================================================
total 2 14.42 721.56 46550 190531 205 100
Misses in library cache during parse: 1
Optimizer goal: CHOOSE
Parsing user id: 173 (recursive depth: 1)
Elapsed times include waiting on following events:
Event waited on Times waited Max. Wait Total Waited
===========================================
db file sequential read 46544 0.49 632.12
db file scattered read 1 0.00 0.00
my select statement
SELECT cst.customer_id> ,DECODE(COUNT(cr.deposit_date), 0, 0, ROUND(SUM(cr.deposit_date - ps.trx_date) / COUNT(cr.deposit_date))) avgdays
> ,DECODE(COUNT(cr.deposit_date), 0, 0, ROUND(SUM(cr.deposit_date - ps.due_date) / COUNT(cr.deposit_date))) avgdayslate
> ,NVL(SUM(DECODE(SIGN(cr.deposit_date - ps.due_date),1, 1, 0)), 0) newlate
> ,NVL(SUM( DECODE(SIGN(cr.deposit_date - ps.due_date),1, 0, 1)), 0) newontime
>
> FROM ar_receivable_applications_all ra
> ,ar_cash_receipts_all cr
> ,ar_payment_schedules_all ps
> ,zz_ar_customer_summary_all cst
> WHERE ra.cash_receipt_id = cr.cash_receipt_id
> AND ra.apply_date BETWEEN ADD_MONTHS(SYSDATE, -12) AND SYSDATE
> AND ra.status = 'APP'
> AND ra.display = 'Y'
> AND ra.applied_payment_schedule_id = ps.payment_schedule_id
> AND ps.customer_id = cst.customer_id
> AND NVL(ps.receipt_confirmed_flag,'Y') = 'Y'
> group by cst.customer_id ;
Thanks,
Anu

Answer 1

user653066 wrote:
Hi All
i found huge diffrence between cpu and elapsed time in tkprof. can you please advice me on this issue.
call     count       cpu    elapsed       disk      query    current        rows
================================================================================
Parse        1      0.12       1.36          2         11          0           0
Execute      1     14.30     720.20      46548     190520        205         100
Fetch        0      0.00       0.00          0          0          0           0
================================================================================
total        2     14.42     721.56      46550     190531        205         100
Misses in library cache during parse: 1
Optimizer goal: CHOOSE
Parsing user id: 173     (recursive depth: 1)
Elapsed times include waiting on following events:
Event waited on                      Times waited   Max. Wait  Total Waited
===========================================================================
db file sequential read                     46544        0.49        632.12
db file scattered read                          1        0.00          0.00
SELECT  cst.customer_id
         ,DECODE(COUNT(cr.deposit_date), 0, 0, ROUND(SUM(cr.deposit_date - ps.trx_date) / COUNT(cr.deposit_date))) avgdays
         ,DECODE(COUNT(cr.deposit_date), 0, 0, ROUND(SUM(cr.deposit_date - ps.due_date) / COUNT(cr.deposit_date))) avgdayslate
         ,NVL(SUM(DECODE(SIGN(cr.deposit_date - ps.due_date),1, 1, 0)), 0)  newlate
         ,NVL(SUM( DECODE(SIGN(cr.deposit_date - ps.due_date),1, 0, 1)), 0) newontime
          FROM ar_receivable_applications_all ra
              ,ar_cash_receipts_all           cr
              ,ar_payment_schedules_all       ps
              ,zz_ar_customer_summary_all cst
          WHERE ra.cash_receipt_id                 = cr.cash_receipt_id
          AND   ra.apply_date                BETWEEN ADD_MONTHS(SYSDATE, -12) AND SYSDATE
          AND   ra.status                          = 'APP'
          AND   ra.display                         = 'Y'
          AND   ra.applied_payment_schedule_id     = ps.payment_schedule_id
          AND   ps.customer_id                     = cst.customer_id          
          AND   NVL(ps.receipt_confirmed_flag,'Y') = 'Y'
          group by cst.customer_id    ;           Toon Koppelaars seems to have pinpointed the problem. Where are the 74 seconds unaccounted for seconds (I might have calculated it incorrectly, but I arrived at 88.08 seconds of unaccounted for time: 721.56 total - 1.36 parse - 632.12 db file sequential reads)?
It is interesting that the maximum wait for a single block read reported by TKPROF is 0.49 seconds - this might be an indication of excessive competition for the server's CPU - processes are waiting in the CPU run queue, and therefore not on the CPU. As Toon indicated, 632.12 of the 721.56 seconds were spent waiting for single block reads to complete with 46,544 blocks read. Note also that the query executed at dep=1, and TKPROF may be providing misleading information about what actually happened during those executions. An example of misleading information:
CREATE TABLE T11 (
  C1 NUMBER,
  C2 VARCHAR2(30));
CREATE TABLE T12 (
  C1 NUMBER,
  C2 VARCHAR2(30));
CREATE TABLE T13 (
  C1 NUMBER,
  C2 VARCHAR2(30));
CREATE TABLE T14 (
  C1 NUMBER,
  C2 VARCHAR2(30));
CREATE OR REPLACE TRIGGER HPM_T11 AFTER
INSERT OR DELETE OR UPDATE OF C1 ON T11
REFERENCING OLD AS OLDDATA NEW AS NEWDATA FOR EACH ROW
BEGIN
  IF INSERTING THEN
    INSERT INTO T12
    SELECT
      ROWNUM,
      DBMS_RANDOM.STRING('A',25)
    FROM
      DUAL
    CONNECT BY
      LEVEL <= 100;
  END IF;
END;
CREATE OR REPLACE TRIGGER HPM_T12 AFTER
INSERT OR DELETE OR UPDATE OF C1 ON T12
REFERENCING OLD AS OLDDATA NEW AS NEWDATA FOR EACH ROW
BEGIN
  IF INSERTING THEN
    INSERT INTO T13
    SELECT
      ROWNUM,
      DBMS_RANDOM.STRING('A',25)
    FROM
      DUAL
    CONNECT BY
      LEVEL <= 100;
  END IF;
END;
CREATE OR REPLACE TRIGGER HPM_T13 AFTER
INSERT OR DELETE OR UPDATE OF C1 ON T13
REFERENCING OLD AS OLDDATA NEW AS NEWDATA FOR EACH ROW
BEGIN
  IF INSERTING THEN
    INSERT INTO T14
    SELECT
      ROWNUM,
      DBMS_RANDOM.STRING('A',25)
    FROM
      DUAL
    CONNECT BY
      LEVEL <= 100;
  END IF;
END;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'MY_TEST_FIND_ME2';
ALTER SESSION SET EVENTS '10046 TRACE NAME CONTEXT FOREVER, LEVEL 8';
SET TIMING ON
INSERT INTO T11 VALUES (1,'MY LITTLE TEST CASE');
ALTER SESSION SET EVENTS '10046 TRACE NAME CONTEXT OFF';The partial TKPROF output:
INSERT INTO T11
VALUES
(1,'MY LITTLE TEST CASE')
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.00          0          8          0           0
Execute      1      0.00       0.00          0       9788         29           1
Fetch        0      0.00       0.00          0          0          0           0
total        2      0.00       0.00          0       9796         29           1
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 56 
Rows     Row Source Operation
      0  LOAD TABLE CONVENTIONAL  (cr=9788 pr=7 pw=0 time=0 us)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  SQL*Net message to client                       1        0.00          0.00
  SQL*Net message from client                     1        0.00          0.00
SQL ID : 6asaf110fgaqg
INSERT INTO T12 SELECT ROWNUM, DBMS_RANDOM.STRING('A',25) FROM DUAL CONNECT
  BY LEVEL <= 100
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.04       0.09          0          2        130         100
Fetch        0      0.00       0.00          0          0          0           0
total        2      0.04       0.09          0          2        130         100
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 56     (recursive depth: 1)
Rows     Row Source Operation
      0  LOAD TABLE CONVENTIONAL  (cr=9754 pr=7 pw=0 time=0 us)
    100   COUNT  (cr=0 pr=0 pw=0 time=0 us)
    100    CONNECT BY WITHOUT FILTERING (cr=0 pr=0 pw=0 time=0 us)
      1     FAST DUAL  (cr=0 pr=0 pw=0 time=0 us cost=2 size=0 card=1)
SQL ID : db46bkvy509w4
INSERT INTO T13 SELECT ROWNUM, DBMS_RANDOM.STRING('A',25) FROM DUAL CONNECT
  BY LEVEL <= 100
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.00          0          0          0           0
Execute    100      1.31       1.27          0         93      10634       10000
Fetch        0      0.00       0.00          0          0          0           0
total      101      1.31       1.27          0         93      10634       10000
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 56     (recursive depth: 2)
Rows     Row Source Operation
      0  LOAD TABLE CONVENTIONAL  (cr=164 pr=0 pw=0 time=0 us)
    100   COUNT  (cr=0 pr=0 pw=0 time=0 us)
    100    CONNECT BY WITHOUT FILTERING (cr=0 pr=0 pw=0 time=0 us)
      1     FAST DUAL  (cr=0 pr=0 pw=0 time=0 us cost=2 size=0 card=1)
SQL ID : 6542yyk084rpu
INSERT INTO T14 SELECT ROWNUM, DBMS_RANDOM.STRING('A',25) FROM DUAL CONNECT
  BY LEVEL <= 100
call     count       cpu    elapsed       disk      query    current        rows
Parse        2      0.00       0.00          0          0          0           0
Execute  10001     41.60      41.84          0       8961      52859     1000000
Fetch        0      0.00       0.00          0          0          0           0
total    10003     41.60      41.84          0       8961      52859     1000000
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 56     (recursive depth: 3)
Rows     Row Source Operation
      0  LOAD TABLE CONVENTIONAL  (cr=2 pr=0 pw=0 time=0 us)
    100   COUNT  (cr=0 pr=0 pw=0 time=0 us)
    100    CONNECT BY WITHOUT FILTERING (cr=0 pr=0 pw=0 time=0 us)
      1     FAST DUAL  (cr=0 pr=0 pw=0 time=0 us cost=2 size=0 card=1)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  log file switch completion                      2        0.07          0.07
********************************************************************************In the above note that the "INSERT INTO T11" is reported as completing in 0 seconds, but it actually required roughly 42 seconds - and that would be visible by manually reviewing the resulting trace file. Also note that the log file switch completion wait was not reported for the "INSERT INTO T11" even though it impacted the execution time.
Back to the possibility of CPU starvation causing lost time. Another test with an otherwise idle server, followed by a second test with the same server having 240 other processes fighting for CPU resources (a simulated load).
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'MY_TEST_QUERY_NO_LOAD';
ALTER SESSION SET EVENTS '10046 TRACE NAME CONTEXT FOREVER, LEVEL 8';
SET TIMING ON
SELECT
  COUNT(*)
FROM
  T14;
SELECT
  SYSDATE
FROM
  DUAL;
SQL> SELECT
  2    COUNT(*)
  3  FROM
  4    T14;
  COUNT(*)
   1000000
Elapsed: 00:00:01.37With no load the COUNT(*) completed in 1.37 seconds. The TKPROF output looks like this:
SQL ID : gy8nw9xzyg3bj
SELECT /* OPT_DYN_SAMP */ /*+ ALL_ROWS IGNORE_WHERE_CLAUSE
  NO_PARALLEL(SAMPLESUB) opt_param('parallel_execution_enabled', 'false')
  NO_PARALLEL_INDEX(SAMPLESUB) NO_SQL_TUNE */ NVL(SUM(C1),:"SYS_B_0"),
  NVL(SUM(C2),:"SYS_B_1")
FROM
(SELECT /*+ NO_PARALLEL("T14") FULL("T14") NO_PARALLEL_INDEX("T14") */
  :"SYS_B_2" AS C1, :"SYS_B_3" AS C2 FROM "T14" SAMPLE BLOCK (:"SYS_B_4" ,
  :"SYS_B_5") SEED (:"SYS_B_6") "T14") SAMPLESUB
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        1      0.01       0.84        523        172          1           1
total        3      0.01       0.84        523        172          1           1
Misses in library cache during parse: 1
Misses in library cache during execute: 1
Optimizer mode: ALL_ROWS
Parsing user id: 56     (recursive depth: 1)
Rows     Row Source Operation
      1  SORT AGGREGATE (cr=172 pr=523 pw=0 time=0 us)
   8733   TABLE ACCESS SAMPLE T14 (cr=172 pr=523 pw=0 time=0 us cost=2 size=12 card=1)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  db file sequential read                         3        0.02          0.04
  db file parallel read                           1        0.31          0.31
  db file scattered read                         52        0.03          0.47
SQL ID : 96g93hntrzjtr
select /*+ rule */ bucket_cnt, row_cnt, cache_cnt, null_cnt, timestamp#,
  sample_size, minimum, maximum, distcnt, lowval, hival, density, col#,
  spare1, spare2, avgcln
from
hist_head$ where obj#=:1 and intcol#=:2
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        1      0.00       0.06          2          2          0           0
total        3      0.00       0.06          2          2          0           0
Misses in library cache during parse: 0
Optimizer mode: RULE
Parsing user id: SYS   (recursive depth: 2)
Rows     Row Source Operation
      0  TABLE ACCESS BY INDEX ROWID HIST_HEAD$ (cr=2 pr=2 pw=0 time=0 us)
      0   INDEX RANGE SCAN I_HH_OBJ#_INTCOL# (cr=2 pr=2 pw=0 time=0 us)(object id 413)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  db file sequential read                         2        0.02          0.04
SELECT
  COUNT(*)
FROM
  T14
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.00          1          1          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        2      0.03       0.43       6558       6983          0           1
total        4      0.03       0.44       6559       6984          0           1
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 56 
Rows     Row Source Operation
      1  SORT AGGREGATE (cr=6983 pr=6558 pw=0 time=0 us)
1000000   TABLE ACCESS FULL T14 (cr=6983 pr=6558 pw=0 time=0 us cost=1916 size=0 card=976987)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  db file sequential read                         1        0.02          0.02
  SQL*Net message to client                       2        0.00          0.00
  db file scattered read                        111        0.02          0.38
  SQL*Net message from client                     2        0.00          0.00Note that TKPROF reported that it only required 0.44 seconds for the query to execute while the SQL*Plus timing indicate that it required 1.37 seconds for the SQL statement to execute. The SQL optimization (parse) with dynamic sampling query accounted for the remaining time, yet TKPROF provided no indication that this was the case.
Now the query with 240 other processes competing for CPU time:
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'MY_TEST_QUERY_WITH_LOAD';
SELECT COUNT(*) FROM T14;
SELECT
  SYSDATE
FROM
  DUAL;
SQL> SELECT COUNT(*) FROM T14;
  COUNT(*)
   1000000
Elapsed: 00:00:59.03The query this time required just over 59 seconds. The TKPROF output:
SQL ID : gy8nw9xzyg3bj
SELECT /* OPT_DYN_SAMP */ /*+ ALL_ROWS IGNORE_WHERE_CLAUSE
  NO_PARALLEL(SAMPLESUB) opt_param('parallel_execution_enabled', 'false')
  NO_PARALLEL_INDEX(SAMPLESUB) NO_SQL_TUNE */ NVL(SUM(C1),:"SYS_B_0"),
  NVL(SUM(C2),:"SYS_B_1")
FROM
(SELECT /*+ NO_PARALLEL("T14") FULL("T14") NO_PARALLEL_INDEX("T14") */
  :"SYS_B_2" AS C1, :"SYS_B_3" AS C2 FROM "T14" SAMPLE BLOCK (:"SYS_B_4" ,
  :"SYS_B_5") SEED (:"SYS_B_6") "T14") SAMPLESUB
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        1      0.00       0.28        423         69          0           1
total        3      0.00       0.28        423         69          0           1
Misses in library cache during parse: 0
Optimizer mode: ALL_ROWS
Parsing user id: 56     (recursive depth: 1)
Rows     Row Source Operation
      1  SORT AGGREGATE (cr=69 pr=423 pw=0 time=0 us)
   8733   TABLE ACCESS SAMPLE T14 (cr=69 pr=423 pw=0 time=0 us cost=2 size=12 card=1)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  db file scattered read                         54        0.01          0.27
  db file sequential read                         2        0.00          0.00
SQL ID : 7h04kxpa13w1x
SELECT COUNT(*)
FROM
T14
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.00       0.03          1          1          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        2      0.06      58.71       6551       6983          0           1
total        4      0.06      58.74       6552       6984          0           1
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 56 
Rows     Row Source Operation
      1  SORT AGGREGATE (cr=6983 pr=6551 pw=0 time=0 us)
1000000   TABLE ACCESS FULL T14 (cr=6983 pr=6551 pw=0 time=0 us cost=1916 size=0 card=976987)
Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  db file sequential read                         1        0.02          0.02
  SQL*Net message to client                       2        0.00          0.00
  db file scattered read                        110        1.54         58.59
  SQL*Net message from client                     1        0.00          0.00Note in the above that the max wait for the db file scattered read is 1.54 seconds due to the extra CPU competition - about 3 times longer than your max wait for a single block read. On your database platform with single block reads, it might be possible that the time in the CPU run queue is not always counted in the db file sequential read wait time or the CPU wait time - what if your operating system is slow at returning timing information to the database instance due to CPU saturation - this might explain the 74 (or 88) lost seconds.
Charles Hooper
IT Manager/Oracle DBA
K&M Machine-Fabricating, Inc.
Edited by: Charles Hooper on Aug 28, 2009 10:26 AM
Fixing formatting problems