Transcript Document

ADABAS 7 PERFORMANCE
CHECKLIST
Jim Poole
October, 2001
1
TOPICS
A. NEW WITH ADABAS 7.
B. THRUPUT ISSUES
C. CPU ISSUES
D. IO ISSUES
2
A. NEW WITH ADABAS 7.
WHY FASTER?
1. LBP ACCESS VIA HASHING
10:1 REDUCTION IN INSTRUCTIONS
IN THE LAB: ABOUT 16% CPU IMPROVEMENT.
2. LFIOP ACCESS VIA HASHING.
3. LFIOP AS A BACKUP LBP.
4. BETTER IMPORTANCE CHAIN DETERMINATION.
5. ASYTVS=YES.
3
1/
3/
0
1/ 1
5/
0
1/ 1
7/
0
1/ 1
9/
1/ 01
11
1/ /01
13
1/ /01
15
1/ /01
17
/
1/ 01
19
1/ /01
21
1/ /01
23
1/ /01
25
/
1/ 01
27
1/ /01
29
1/ /01
31
/0
2/ 1
2/
0
2/ 1
4/
0
2/ 1
6/
0
2/ 1
8/
2/ 01
10
2/ /01
12
/
2/ 01
14
2/ /01
16
2/ /01
18
2/ /01
20
/
2/ 01
22
2/ /01
24
2/ /01
26
2/ /01
28
/0
3/ 1
2/
0
3/ 1
4/
0
3/ 1
6/
0
3/ 1
8/
3/ 01
10
3/ /01
12
3/ /01
14
3/ /01
16
/0
1
seconds
ADABAS 7 CPU / CMDTIME DIFFERENCES
MEAN CMD TIME (V7: 02/18)
0.0001
0.00009
0.00008
0.00007
0.00006
0.00005
V7
0.00004
0.00003
0.00002
0.00001
0
4
A. NEW WITH ADABAS 7.
FORWARD INDEX COMPRESSION (FIC)
1. SAVE INDEX SPACE:
46 TO 54%
2.
SAVE OVERALL DATABASE SPACE: 10 TO 13 %
3.
SAVE ON NEW DASD FOR GROWTH: 10 TO 13 %
4.
INCREASE POTENTIAL and EFFICIENCY FOR NI CACHE:
CACHE UI AC
(eg. 15K 16K BLOCKS)
CACHE NI ??? (eg. 1700K BLOCKS)
CACHE DS ?? (eg.
5. REDUCE IO FOR L3/L6, L9, Sx:
400K BLOCKS)
DATABASE OVERALL
BATCH JOBS DOING LONG L3/L6
6. IMPROVED BUFFER EFFICIENCY.
7.
FASTER UTILITIES: ADASAV, ADAORD, ETC.
5
2/
15
/
2/ 01
16
/
2/ 01
17
/
2/ 01
18
/
2/ 01
19
/
2/ 01
20
/
2/ 01
21
/
2/ 01
22
/
2/ 01
23
/
2/ 01
24
/
2/ 01
25
/
2/ 01
26
/
2/ 01
27
/
2/ 01
28
/0
3/ 1
1/
0
3/ 1
2/
0
3/ 1
3/
0
3/ 1
4/
0
3/ 1
5/
0
3/ 1
6/
0
3/ 1
7/
0
3/ 1
8/
0
3/ 1
9/
3/ 01
10
/
3/ 01
11
/
3/ 01
12
/
3/ 01
13
/
3/ 01
14
/
3/ 01
15
/
3/ 01
16
/
3/ 01
17
/
3/ 01
18
/
3/ 01
19
/
3/ 01
20
/
3/ 01
21
/0
1
seconds
A. NEW WITH ADABAS 7. (FIC)
JOB NONAME RUN TIME
6000
5000
4000
FIC->
3000
2000
1000
0
6
A. NEW WITH ADABAS 7. (FIC)
? ANY CPU OVERHEAD ON READS ?
? ANY CPU OVERHEAD ON UPDATES?
! NONE !
FIC EVERYTHING.
7
RUNS
8
534
521
508
495
482
469
456
443
430
417
404
391
378
365
352
339
326
313
300
287
274
261
248
235
222
209
196
183
170
157
144
131
118
105
92
79
66
53
40
27
14
1
TIME (MINUTES)
A. NEW WITH ADABAS 7. ADARES REGENGERATE MTR=YES
30
25
20
15
MTR=NO ->
10
MTR=YES ->
5
0
A. NEW WITH ADABAS 7. ONLINE INV, ORD
ONLINE INVERT:
1. RUN TIME: SAME AS BATCH (IF NO DATABASE ACTIVITY)
2. USES A TAD MORE NI SPACE THAN BATCH (+2%)
3. DURING ADASAV:
WATCH THOSE PLOG IO’S
4. ADARES REGENERATES NICELY.
ONLINE ORD:
1. RUN TIME: 5X BATCH (IF NO DATABASE ACTIVITY)
2. RESPECTS PAD MORE THAN BATCH RESPECTS PAD
3. DURING ADASAV:
WATCH THOSE PLOG IO’S
4. ADARES REGENERATES DATA NICELY.
9
B.THRUPUT ISSUES. OVERVIEW.
TIME 1
USER
TIME 2 TIME 3
CQ
THREAD
TIME 4
USER
10
B.THRUPUT ISSUES. TOOLS.
1. ADADBS DCQ
2. ADADBS ECBS (OFF CLOG)
3. THREAD COUNT (SHUTDOWN, DTH, etc)
4. ENQ/DUR (ADARUN CL=4/5)
5. SYNS-5B, ADADBS REFRESHSTATS
6. IS THE LOAD IMPACTING PERFORMANCE?
11
B.THRUPUT ISSUES. TOOLS.
TIME 1
USER
TIME 2 TIME 3
CQ
TIME 4
THREAD
X’18’
USER
X’28’
1. ADADBS DCQ
USER
JOB
FLAG
IUBL
---------------------------------------C3C9C3E2C2F1C8E3 (CICSB1HT) CIPRCSBC 28
2927
C3C9C3E2C2F1D2D9 (CICSB1KR) CIPRCSBA 28
2927
C3C9C3E2C2F1C2D8 (CICSB1BQ) CIPRCSBC 28
367
C3C9C3E2C2F1C1F8 (CICSB1A8) CIPRCSBB 18
304
B65C11176BA318C3 (:*..,t.C) P1BXK0AR 28
2530
C3C9C3E2C2F17CD9 (CICSB1@R) CIPRCSBB 18
2927
C3C9C3E2C2F1C2D8 (CICSB1BQ) CIPRCSBC 28
367
C3C9C3F7F1F3F4F4 (CIC71344) CIPRCSBH 18
634
C3C9C3E2C2F15F7B (CICSB1¬#) CIPRCSBA 18
2927
C3C9C3E2C2F1E55E (CICSB1V;) CIPRCSBC 28
380
12
B.THRUPUT ISSUES. TOOLS.
TIME 1
USER
TIME 2 TIME 3
CQ
THREAD
TIME 4
USER
ECB COUNT
2. ADADBS ECBS (OFF CLOG)
TIME
-----165402
165501
170001
170501
171001
171501
172001
CMD
COUNT
--------291,784
1,578,348
1,949,641
2,171,444
2,018,170
2,178,302
116,406
MEAN
ENQTIME
SECS
------0.00003
0.00003
0.00002
0.00002
0.00001
0.00002
0.00002
MEAN
DURAT
SECS
------0.00063
0.00047
0.00051
0.00038
0.00033
0.00036
0.00052
MEAN MEAN MEAN
CMD PST TOT
TIME ECB
IO
------- ---- ---0.00006 5.8 0.06
0.00006 5.3 0.03
0.00005 5.1 0.05
0.00004 4.3 0.04
0.00004 3.7 0.04
0.00004 4.7 0.03
0.00005 7.1 0.03
MAX
PST MAX
ECB THD
--- --24
9
21 10
43 15
20 18
21 11
34 14
18
8
13
B.THRUPUT ISSUES. TOOLS.
TIME 1
USER
TIME 2 TIME 3
CQ
THREAD
TIME 4
USER
THREAD
STATUS
USAGE
THRD CNT -----------------------1
AA 118574503
3. THREAD COUNT (SHUTDOWN, DTH, etc)
2
WI 108594864
3
WI
79809114
4
WI
53275850
5
WI
30081855
6
UU
8260981
7
UU
3245993
8
UU
1740369
? 90 TO 95 % OF
9
UU
1146354
COMMANDS ?
10
UU
788271
11
UU
523565
12
UU
287192
13
UU
205306
14
UU
159414
15
UU
126467
14
16
UU
99034
B.THRUPUT ISSUES. TOOLS.
TIME 1
USER
TIME 2 TIME 3
CQ
CL=4
THREAD
ENQ
TIME 4
USER
DUR
ENQ
DUR
CL=5
4. ENQ/DUR (ADARUN CL=4/5)
TIME
-----165402
165501
170001
170501
171001
171501
172001
CMD
COUNT
--------291,784
1,578,348
1,949,641
2,171,444
2,018,170
2,178,302
116,406
MEAN
ENQTIME
SECS
------0.00003
0.00003
0.00002
0.00002
0.00001
0.00002
0.00002
MEAN
DURAT
SECS
------0.00063
0.00047
0.00051
0.00038
0.00033
0.00036
0.00052
MEAN MEAN MEAN
CMD PST TOT
TIME ECB
IO
------- ---- ---0.00006 5.8 0.06
0.00006 5.3 0.03
0.00005 5.1 0.05
0.00004 4.3 0.04
0.00004 3.7 0.04
0.00004 4.7 0.03
0.00005 7.1 0.03
MAX
PST MAX
ECB THD
--- --24
9
21 10
43 15
20 18
21 11
34 14
18
8
15
B.THRUPUT ISSUES. TOOLS.
5. SYNS-5B, ADADBS REFRESHSTATS
CHECKPOINT CREATED BY:
a. NUCLEUS SHUTDOWN (eg. ADAEND)
b. ADADBS REFRESHSTATS
ALL / CMDUSAGE / COUNTERS / FILEUSAGE
POOLUSAGE / THREADUSAGE
REPORT ON CHECKPOINT:
a. ADAREP CPEXLIST,OFFSET=nn
b. DDPRINT OFF NUC (IDENTICAL TO SHUTDOWN STATS)
16
B.THRUPUT ISSUES. TOOLS. 5. SYNS-5B, REFRESHSTATS
SYNS 5B
2001-09-09 02:35:52
REFRESH STATISTICS
THREAD
1: 983037373 COMMANDS
THREAD
2: 869948405 COMMANDS
THREAD
3: 566715558 COMMANDS
: :
: :
: :
REMOTE LOGICAL CALLS
186217341
REMOTE PHYSICAL CALLS
0
LOCAL LOGICAL CALLS
3147407681
LOCAL PHYSICAL CALLS
0
FORMAT TRANSLATIONS
127578298
FORMAT OVERWRITES
42929
AUTORESTARTS
162896
PLOG SWITCHES
145
BUFFER FLUSHES
5353
THROWBACKS ISN PROBS
3536
THROWBACKS SPACE PROBS
0
17
B.THRUPUT ISSUES. TOOLS. 5. SYNS-5B, REFRESHSTATS
HIGH WATER MARKS:
COMMAND QUEUE
31296
HOLD QUEUE
112084
FORMAT POOL
2999916
TBI POOL
4288
TBS POOL
76496
USER QUEUE
196444
WORK POOL
636832
SECURITY POOL
0
UQFLIST POOL
53056
UNIQUE DE POOL
17
XID POOL
0
A1/4 COMMANDS
14749218
BT
COMMANDS
164487
:
: :
: :
18
B.THRUPUT ISSUES. TOOLS.
6. IS THE LOAD IMPACTING PERFORMANCE?
SUM OF DURATION
ELAPSED TIME
IF > 1, THEN YES, THERE ARE PERFORMANCE IMPACTS.
IF < 1, THEN NO, WHO CARES.
19
B.THRUPUT ISSUES. SOLUTIONS.
1. WAIT-ON-HOLD.
2. COMPLEX COMMAND WAIT.
3. UPDATE COMMAND WAIT.
4. CLOG IO.
5. UEX4 AND COMMAND LOG TOOL.
6. EXCESSIVE COMMAND LOG REPORTING.
7. BAD UEX1/2.
8. PATH LENGTH.
9. SYSTEM PRIORITIES.
10. EXCESSIVE BATCH JOBS.
11. EXCESSIVE MULTI/PREFETCH JOBS.
12. EXCESSIVE BAD SEARCHES.
13. EXCESSIVE NON-DESCRIPTOR SEARCHES.
14. EXCESSIVE SX/L3 ON EXPANDED FILES.
15. USER WORKLOAD IMPROVEMENTS.
16. NEW DATABASE VS. SMP.
17. SEE CPU ISSUES.
18. SEE IO ISSUES.
20
B.THRUPUT ISSUES. SOLUTIONS.
1. WAIT-ON-HOLD.
L4,L5,L6,S4,HI
vs.
L1,L2,L3,S1
HOLDERS: SUMMARIZE(SUBTOT=OFF)
COUNT
MEAN(ENQ-TIME) (PF=N1.5)
MEAN(DURATION) (PF=N2.5)
MAX(DURATION)
MEAN(CMD-TIME) (PF=N1.5)
MEAN(TOTAL-IO) (PF=N1.2)
MEAN(ECBS)
(PF=N2.2)
BY CMD
***
INTERVAL N HOURS
WHERE
CMD(1) = 'L'
TITLE = ‘HOLD VS NOT HOLD’
INSIGHT-LINES = 100
**
PRINT TO SUMM2
;
21
B.THRUPUT ISSUES. SOLUTIONS. 1. WAIT-ON-HOLD.
ANALYSYS OF HOLD OPTION
C
MEAN
MEAN
MAX
MEAN MEAN MEAN
M
CMD ENQTIME
DURAT
DURAT
CMD TOT
PST
D
COUNT
SECS
SECS
SECS
TIME
IO
ECB
-- ------- ------- ------- ------- ------- ---- ----L1 200,659 0.00003 0.00037
0.286 0.00003 0.04 5.07
L4
3,286 0.00003 0.43218
9.456 0.00007 0.03 6.58
L2 204,753 0.00003 0.00022
0.061 0.00003 0.02
4.19
L3 542,935 0.00003 0.00056
L6
1,296 0.00003 0.00268
0.672 0.00006 0.05
0.138 0.00026 0.21
5.31
6.22
22
B.THRUPUT ISSUES. SOLUTIONS. 1. WAIT-ON-HOLD.
LONGHOLD: EXTRACT(ML=500000)
TIME
*
JOBNAME
TP-TRANS-NAME (HD='TRAN','NAME')
*
PROGRAM NATURAL-LOGON
NAT-PROGRAM
CID
(PF=H8)
FILE
(PF=N3)
COMMAND-CODE
COPT1
DURATION
(PF=N3.3)
*
ENQ-TIME
(PF=N1.5)
CMD-TIME
(PF=N1.5)
ISN
(PF=N8)
TOTAL-IO
(PF=N4)
WHERE
(CMD = 'L4' OR CMD = 'L6‘)
AND DURATION > 5
INSIGHT-LINES = 300
23
**
PRINT TO ADHOC2;
B.THRUPUT ISSUES. SOLUTIONS. 1. WAIT-ON-HOLD.
C
TRAN NATURAL
FIL M
TIME NAME PROGRAM
CID
NUM D
110948 NATP LQNWCTN2 47200201 95 L4
ISN
1,973,382
C
O
1
.
DURAT
CMD
SECS
TIME
9.456 0.00016
TOT
IO
0
NUC SHUTDOWN DDPRINT or SYNS-5B CHECKPOINT
THROWBACKS ISN PROBS
3536
MEASURED OVER TIME.
24
B.THRUPUT ISSUES. SOLUTIONS. 1. WAIT-ON-HOLD.
SOLUTIONS:
1. CB COP1=‘R’
RSP CD 145
2. NATURAL WH=OFF (SECURITY), TRAP FOR NON-0 RSP CD.
NOTROPT: SUMMARIZE(SUBTOT=OFF)
COUNT
MIN(CID)
(PF=H8)
MAX(CID)
(PF=H8)
MIN(FILE)
(PF=N3)
MAX(FILE)
(PF=N3)
BY NAT-PROGRAM
WHERE
(CMD = 'L4' OR CMD = 'L6')
AND COPT1 NE 'R'
AND NAT-PROGRAM > ' '
*
AND CID NE H'00000000'
INSIGHT-LINES = 100
**
PRINT TO SUMM2;
25
B.THRUPUT ISSUES. SOLUTIONS. 1. WAIT-ON-HOLD.
NATURAL
CMD
MIN
PROGRAM
COUNT
CID
-------- ---------- -------LQPBKY18
1 43050101
LQCCVDN1
36 07500201
LQVENKP3
13 07500201
LQVXAPP1
23 33450101
MIN MAX
MAX
FIL FIL
CID
NUM NUM
-------- --- --43050101 182 182
10900201 76 78
07500201 182 182
34100101 131 169
? WHEN IS ‘R’ OPTION NOT GOOD?
? BATCH ?
? BACKGROUND ?
? ?
26
B.THRUPUT ISSUES. SOLUTIONS.
2. COMPLEX COMMAND WAIT.
USER
CQ
?
THREAD
COMPLEX CMD IF :
1.
2.
CMD = S2, S8, S9
SBL > 9
COMPLEX COMMANDS NOT SELECTED
FROM CQ TO THREAD IF:
1.
2.
3.
LESS THAN 50% WORK-2 AVAILABLE.
LESS THAN LS AVAILABLE IN LWP
AT LEAST ONE COMMAND WAITING
ON LWP SPACE
27
B.THRUPUT ISSUES. SOLUTIONS. 2. COMPLEX COMMAND WAIT.
CMDTYPE: SUMMARIZE (SUBTOT=ON)
COUNT
MEAN(ENQ-TIME) (PF=N1.5)
MEAN(DURATION) (PF=N1.5)
MEAN(CMD-TIME) (PF=N1.5)
MEAN(ECBS)
(PF=N2.1)
MEAN(TOTAL-IO) (PF=N1.2)
**
BY JOBNAME
BY COMMAND-TYPE
INTERVAL 15 MINUTES
INSIGHT-LINES 100
**
PRINT TO SUMM1
;
28
B.THRUPUT ISSUES. SOLUTIONS. 2. COMPLEX COMMAND WAIT.
T
MEAN
MEAN
MEAN MEAN MEAN
Y
CMD ENQTIME
DURAT
CMD PST TOT
P
COUNT
SECS
SECS
TIME ECB
IO
1. CL=5 C
435 0.00002 0.00139 0.00013 7.8 0.04
S 335,188 0.00002 0.00106 0.00005 5.4 0.08
U 28,500 0.00002 0.00643 0.00050 8.6 0.44
T
MEAN
MEAN
MEAN MEAN MEAN
Y
CMD ENQTIME
DURAT
CMD PST TOT
COUNT
SECS
SECS
TIME ECB
IO
2. CL=5 P
C
435 0.00002 0.54339 0.00013 7.8 0.04
S 335,188 0.00002 0.00106 0.00005 5.4 0.08
U 28,500 0.00002 0.00643 0.00050 8.6 0.44
T
MEAN
MEAN
MEAN MEAN MEAN
Y
CMD ENQTIME
DURAT
CMD PST TOT
P
COUNT
SECS
SECS
TIME ECB
IO
3. CL=4
C
435 0.54307 0.00131 0.00013 7.8 0.04
S 335,188 0.00008 0.00100 0.00005 5.4 0.08
29
U 28,500 0.00009 0.00636 0.00050 8.6 0.44
B.THRUPUT ISSUES. SOLUTIONS. 2. COMPLEX COMMAND WAIT.
NUC SHUTDOWN DDPRINT or SYNS-5B CHECKPOINT
THROWBACKS SPACE PROBS
190
30
B.THRUPUT ISSUES. SOLUTIONS. 2. COMPLEX COMMAND WAIT.
a. WORK-2 SOLUTION
ADARUN LWKP2=
! USE !
MIN (<16M REC FILES):
24,000 (3390)
10,000 (8391)
LWKP2 = ( S x 2 x TOPISN ) / ( ( WB – 16) / 4 )
WB
16
4
TOPISN
2
S
=
=
=
=
=
=
5724 or 13682
BLOCK HEADER
4b PER ISN
OF LARGEST FILE FOR SX CMDS
ALWAYS ENSURE 50% AVAILABLE
PEAK NO. OF CONCURRENT COMPLEX COMMANDS.
%Sx TIMES USED THREADS (?TIMES 2?)
31
B.THRUPUT ISSUES. SOLUTIONS. 2. COMPLEX COMMAND WAIT
b. WORK-POOL SOLUTION
ADARUN LWP/LS=
LWP = LS * 3
= LS + (HI-WATER BYTES FROM DDPRINT)
= LS * S (WHERE S = 2 * (%Sx) * NT)
= LS * NT
32
B.THRUPUT ISSUES. SOLUTIONS.
3. UPDATE COMMAND WAIT.
USER
CQ
?
THREAD
UPDATE CMD IF :
CMD = A1,A4,E1,E4,
N1,N2,
ET,BT,OP,CL
UPDATE COMMANDS NOT SELECTED
FROM CQ TO THREAD IF:
1.
LFIOP=0 AND BUFFER FLUSH,
WORK WRITE, OR PLOG WRITE.
2.
ET-SYNC:
ADAEND
ADASAV SYN2
OP CMD SYNCC
33
B.THRUPUT ISSUES. SOLUTIONS. 3. UPDATE COMMAND WAIT.
T
MEAN
MEAN
MEAN MEAN MEAN
Y
CMD ENQTIME
DURAT
CMD PST TOT
P
COUNT
SECS
SECS
TIME ECB
IO
1. CL=5 C
435 0.00002 0.00139 0.00013 7.8 0.04
S 335,188 0.00002 0.00106 0.00005 5.4 0.08
U 28,500 0.00002 8.12643 0.00050 8.6 0.44
T
MEAN
MEAN
MEAN MEAN MEAN
Y
CMD ENQTIME
DURAT
CMD PST TOT
P
COUNT
SECS
SECS
TIME ECB
IO
2. CL=4 C
435 0.00009 0.00130 0.00013 7.8 0.04
S 335,188 0.00008 0.00098 0.00005 5.4 0.08
U 28,500 8.11002 0.00643 0.00050 8.6 0.44
34
B.THRUPUT ISSUES. SOLUTIONS. 3. UPDATE COMMAND WAIT.
a. LFIOP SOLUTION
ADARUN LFIOP= 1 or n
! USE !
b. ADASAV SOLUTION
NOTE UPDATE LOAD VERSUS ADASAV TIME
ADASAV TTSYN=120 ?
35
B.THRUPUT ISSUES. SOLUTIONS. 3. UPDATE COMMAND WAIT.
* USE ADARUN CL=4
MCISYN2: SUMMARIZE
COUNT
MEAN(ENQ-TIME)
****
MIN(ENQ-TIME)
****
MAX(ENQ-TIME)
MIN(TIME)
MAX(TIME)
BY ENQ-TIME (BY 5)
WHERE
JOBNAME2(2) = 'CI'
AND TIME GE '204500‘
AND TIME LE '235959'
AND ENQ-TIME LE 240
INTERVAL 15 MINUTE
TITLE='TIME TO SYN2 TO ET-SYNC'
PRINT TO ADHOC1;
36
B.THRUPUT ISSUES. SOLUTIONS.
4. CLOG IO.
! ELIMINATE OR REDUCE !
EACH CLOG IO IS A WAIT IO !
DATABASE XYZ
FULL COMMAND LOGGING
200 COMMANDS PER SEC
MEAN ECBS = 11.3
NOLOGGING
MEAN ECBS =
7.0
1400 CMDS PER SEC
FULL CLOGGING
CLOG BLKSIZE = 5064
Cloglayout
= 5
LRECL
= 172
RECS PER BLK = 29
IO’S PER SEC = 48.3
MEAN IO TIME = .004
48.3 X .004 = .19 SEC
.19 / 1 = 19% WAITTIME
37
B.THRUPUT ISSUES. SOLUTIONS.
5. UEX4 AND COMMAND LOG TOOL.
HOMEGROWN
TRIM: PART OF NUC
REVIEW: SUBTASK OR HUB
APAS/INSIGHT: SUBTASK
38
B.THRUPUT ISSUES. SOLUTIONS.
6. EXCESSIVE COMMAND LOG REPORTING.
???
EVERY HOUR,
DB EXPERIENCING
SLOWDOWN
???
SUMMARY REPORT:
COUNT, DURATION,ETC.
BY JOBNAME
BY NAT-PROGRAM
BY FILE
BY CMD
INTERVAL 1 HOUR
39
B.THRUPUT ISSUES. SOLUTIONS.
7. BAD UEX1/2.
DATABASE IS SINGLE-THREADED AT THIS TIME.
40
B.THRUPUT ISSUES. SOLUTIONS.
8. PATH LENGTH.
SAME LPAR, DIFFERENT LPAR, REMOTE
JOB NAME
-------CICSPROD
DB00SAME
DB00LPAR
DB00VTAM
CMD
COUNT
-----1,416
1,002
1,002
1,002
MEAN
ENQTIME
SECS
------0.00002
0.00003
0.00081
0.18632
MEAN
DURAT
SECS
------0.00152
0.00006
0.00009
0.00014
MEAN MEAN MEAN
CMD PST TOT
TIME ECB
IO
------- ---- ---0.00027 2.0 0.89
0.00001 1.9 0.02
0.00002 1.9 0.02
0.00002 2.4 0.05
41
B.THRUPUT ISSUES. SOLUTIONS.
9. SYSTEM PRIORITIES.
PRE - WLM
1. ADABAS
2. TP
3. BATCH
WLM
1. VELOCITY
HIGH
2. SYSSTC
? CPU WAIT ?
? ENQ-TIME ?
? CMD-TIME ?
42
B.THRUPUT ISSUES. SOLUTIONS.
10. EXCESSIVE BATCH JOBS.
a. EACH BATCH JOB TENDS TO DOMINATE ONE MORE
THREAD
11. EXCESSIVE MULTI/PREFETCH JOBS.
a. SHOULD RUN NON-PEAK TIMES.
b. SHOULD RUN FEWER THAN 2-3 CONCURRENTLY.
43
B.THRUPUT ISSUES. SOLUTIONS. 10,11. EXCESSIVE JOBS.
MEAN ECBS VS BATCH JOBS
35
30
25
20
15
10
5
ECBS
281
44
274
267
260
253
246
239
232
225
218
BATCH JOBS
COUNT
211
204
197
190
183
176
169
162
155
148
141
134
127
ECBS
120
113
106
99
92
85
78
71
64
57
50
43
36
29
22
15
8
1
0
B.THRUPUT ISSUES. SOLUTIONS. 10,11. EXCESSIVE JOBS.
DECLARE FETCH (A3) VALOF
IF COPT1
= 'M' RESULTIS 'YES'
ELSEIF COPT1 = 'P' RESULTIS 'YES'
ELSEIF COPT1 = 'O' RESULTIS 'YES'
ELSE
RESULTIS '
'
IFEND;
DBUSE: SUMMARIZE(GTT=YES)
COUNT
MEAN(ENQ-TIME) (PF=N1.5)
MEAN(DURATION) (PF=N1.5)
MEAN(CMD-TIME) (PF=N1.5)
MEAN(ECBS)
(PF=N2.1)
MEAN(TOTAL-IO) (PF=N1.2)
MAX(FETCH)
(HD='FTCH')
BY JOBNAME
INTERVAL 15 MINUTES
INSIGHT-LINES=300;
45
B.THRUPUT ISSUES. SOLUTIONS. 10,11. EXCESSIVE JOBS.
JOB NAME
CICSPRDA
CICSPRDB
CICSPRDC
CICSPRDD
CICSPRDE
CICSPRDF
CICSPRDG
CICSPRDH
CICSPRDU
JOB00001
JOB00002
JOB00003
JOB00004
JOB00005
JOB00006
========
CMD
COUNT
39
2,217
10,231
524
1,055
274
2,246
311
12
6,660
5,815
5,497
3,892
3,127
15,833
57,733
MEAN
ENQTIME
SECS
0.00002
0.00003
0.00002
0.00003
0.00003
0.00002
0.00002
0.00003
0.00118
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00001
MEAN
DURAT
SECS
0.00045
0.00037
0.00093
0.00029
0.00070
0.00041
0.00172
0.00059
0.00060
0.00104
0.00118
0.00130
0.00032
0.00031
0.00027
0.00074
MEAN MEAN MEAN
CMD PST TOT
TIME ECB
IO FTCH
0.00012 10.4 0.18
0.00007 9.3 0.04
0.00006 9.2 0.14
0.00006 9.3 0.04
0.00009 9.7 0.10
0.00005 9.7 0.02
0.00021 8.9 0.34
0.00005 8.9 0.05
0.00007 10.8 0.17
0.00005 8.5 0.18
0.00004 8.4 0.07
0.00004 8.4 0.09
0.00003 9.4 0.02
0.00002 9.5 0.02
0.00002 8.5 0.03
0.00005 8.8 0.09 46
B.THRUPUT ISSUES. SOLUTIONS. 10,11. EXCESSIVE JOBS.
JOB NAME
-------CICSPRD1
CICSPRD2
CICSPRD3
CICSPRD4
DDBTCHA1
DDBTCHB6
DDBTCHHQ
DDBTCHXY
CMD
COUNT
------5,261
494
267
233
5,158
8
4,407
226
MEAN
ENQTIME
SECS
------0.00003
0.00003
0.00003
0.00003
0.00002
0.00002
0.00003
0.00003
MEAN
DURAT
SECS
------0.00073
0.00030
0.00095
0.00212
0.00109
0.00020
0.00200
0.06774
MEAN MEAN MEAN
CMD PST TOT
TIME ECB
IO FTCH
------- ---- ---- ---0.00008 3.7 0.13
0.00011 4.2 0.03
0.00009 3.1 0.21
0.00009 4.5 0.22
0.00016 4.6 0.09
0.00006 5.5 0.00
0.00015 3.6 1.11
0.00113 7.7 5.65 YES
47
B.THRUPUT ISSUES. SOLUTIONS.
12. EXCESSIVE BAD SEARCHES.
FIND
CLIENTS
WITH
AND
AND
AND
CITY=‘NEW YORK’
NAME=‘SMITH’
AGE>21
WEIGHT=321
48
B.THRUPUT ISSUES. SOLUTIONS. 12. BAD SEARCHES.
HIGHIO:SUMMARIZE(SUBTOT=OFF) HIGHIOE: EXTRACT(ML=1000000)
COUNT
DATE-TIME
MEAN(TOTAL-IO)
COMMAND-CODE
MEAN(DATA-IO)
FILE
MEAN(ISNQ)
JOBNAME
MEAN(DURATION)
CID
(PF=H8)
BY JOBNAME
TP-TRANS-NAME
BY NAT-PROGRAM
NAT-LOGON
BY CID (PF=H8)
NAT-PROGRAM
BY CMD
PROGRAM
WHERE
CICS-TERMID
TOTAL-IO GT 100
ISNQ
OR DURATION GT 1.0
TOTAL-IO
DURATION
*
PRINT TO SUMBAD
WHERE . . . .
INSIGHT-LINES 300;
49
B.THRUPUT ISSUES. SOLUTIONS. 12. BAD SEARCHES.
C
MEAN MEAN
JOB/
M
CMD
TOT DATA
TRANS
PROGRAM
CID
D COUNT
IO
IO
-------- -------- -------- -- ----- ------------NATO
XYVADDP1 01300101 S1
3
178.3
0.0
MEAN
MEAN
ISN
DURAT
QUANT
SECS
-------- ------29,527.0 0.4050
50
B.THRUPUT ISSUES. SOLUTIONS.
13. EXCESSIVE NON-DESCRIPTOR SEARCHES.
a. REPORT ON VIA: WHERE
CMD(1) = ‘S’
AND DATA-IO>1
b. STOP VIA: ADARUN NONDES=NO
51
B.THRUPUT ISSUES. SOLUTIONS.
14. EXCESSIVE SX/L3 ON EXPANDED FILES.
EXPANDED FILE
vs.
4-BYTE ISN FILE
MANAGEABILITY
vs
PERFORMANCE
SPACE SAVINGS
vs
SPACE SAVINGS
(3B vs 4B ISN IN ASSO)
(NO DUPE DE-VALUE
ACROSS FILES)
! UNLESS MANAGEABILITY IS AN ISSUE, USE 4-BYTE ISN FILES !
(vs EXPANDED FILES)
52
B.THRUPUT ISSUES. SOLUTIONS. 14. EXCESS SX/L3, EXP.FILES.
Sx : Eg. NAME=BOB
NON-EXP. FILE
FILE-X NI :
BOB 1,5,8
R.I.L. 1,5,8,9,10,11
(Resultant Isn List)
EACH NI BLOCK: MADE UP OF NI ELEMENTS.
EACH NI ELEMENT: DE-VALUE / COUNT / ISN LIST
53
B.THRUPUT ISSUES. SOLUTIONS. 14. EXCESS SX/L3, EXP.FILES.
Sx : Eg. NAME=BOB
EXP. FILE (4 COMPONENTS)
FILE-W NI :
FILE-X NI :
FILE-Y NI :
BOB 1
BOB 5,8
BOB 9
FILE-Z NI : BOB 10,11
R.I.L. 1,5,8,9,10,11
(Resultant Isn List)
54
B.THRUPUT ISSUES. SOLUTIONS. 14. EXCESS SX/L3, EXP.FILES.
L3/6 THRU FILE
NON-EXP. FILE
FILE-X NI : BART
L3’S:
1
BEA,BETH
2
3
BILL
4
NOTE: 1 TBQ(LQ=) ELEMENT CREATED AND USED.
BOB,BRAD
5
6
55
B.THRUPUT ISSUES. SOLUTIONS. 14. EXCESS SX/L3, EXP.FILES.
L3/6 THRU FILE
EXP. FILE (4 COMPONENTS)
FILE-W NI : BART 1
BETH 5
FILE-X NI : BART 2
BEA
FILE-Y NI : BETH 7
BILL
FILE-Z NI : BEA 4
BOB 11
3
8
BOB
9
BUD 13
6
. . .
BOB 10
. . .
BRAD 12
. . .
BETH
L3’S: FOLLOW THE NUMBERS
NOTE: 4 TBQ(LQ=) ELEMENTS CREATED AND USED.
56
B.THRUPUT ISSUES. SOLUTIONS. 14. EXCESS SX/L3, EXP.FILES.
CUSTOMER EXPERIENCE (SAG-L)
BEFORE: 29 COMPONENT FILES, 3B ISN, 16M RECS @
AFTER:
10 COMPONENT FILES, 4B ISN, 50M RECS @
BEFORE: L3’S DOING UP TO 60 IO’S
AFTER: L3’S DOING UNDER 10 IO’S
BEFORE: BATCH JOB RUNS IN 11 HOURS
AFTER: BATCH JOB RUNS IN 3 HOURS
57
B.THRUPUT ISSUES. SOLUTIONS.
15. USER WORKLOAD IMPROVEMENTS.
a. HISTORICAL – PERIODIC REVIEW OF THE TOP 10%
b. ONLINE
WHEN
WHO
BY
BY
BY
BY
JOBNAME
TP-TRANS-NAME
PROGRAM
NAT-PROGRAM
BY NAT-PROGRAM
WHERE JOBNAME(3)=‘CIC’
AND NAT-PROGRAM > ‘ ‘
PEAK HOURS
DAILY
WEEKLY
INTERVAL 168 HOURS
WHERE TIME GE ‘090000’
AND TIME LE ‘170000’
WHAT
COUNT
DURATION
CMD-TIME
TOTAL-IO
58
B.THRUPUT ISSUES. SOLUTIONS. 15. USER WORKLOAD.
a. HISTORICAL – PERIODIC REVIEW OF THE TOP 10
PROGRAM
LQNSCTNH
LQ2DCTN1
LQLWNTN1
GGLGNUP2
LQLGLCA1
GGLCNEP1
LQ24C2N3
LQNWDGF2
LQLCNTP1
SECUPTER
COUNT
33,455,949
25,876,640
23,581,968
9,548,540
7,254,208
6,529,145
6,420,988
5,486,722
5,270,677
4,956,576
DURATION
5,888.379
7,441.052
10,270.808
2,392.155
9,371.013
1,902.695
1,472.643
2,103.440
1,802.396
1,227.651
TOTALIO
786
396,146
880,891
26,559
1,248,313
49,985
15,476
57,322
46,019
1,473
CMDTIME
826.723
1,228.789
791.377
498.448
495.574
352.686
291.566
382.099
412.284
351.807
SORT BY: COUNT , DURATION , TOTALIO , CMDTIME
REVIEW USER BY: (CID), FILE, CMD
59
B.THRUPUT ISSUES. SOLUTIONS. 15. USER WORKLOAD.
b. ONLINE
QUICK:
*
*
*
*
*
*
*
SUMMARIZE
COUNT
MEAN(ENQ-TIME)
MEAN(DURATION)
MEAN(CMD-TIME)
MEAN(ECBS)
MEAN(TOTAL-IO)
INTERVAL 10 SECONDS
BY JOBNAME 1
BY TP-TRANS-NAME
2
WHERE JOBNAME = 'CICSPRDA'
BY GLOBAL-USER
WHERE JOBNAME = 'CICSPRDA' 3
AND TP-TRANS-NAME = 'NATP'
INSIGHT-LINES=300;
60
B.THRUPUT ISSUES. SOLUTIONS. 15. USER WORKLOAD.
b. ONLINE
BY JOBNAME
JOB NAME
-------CICSPRDA
CICSPRDB
CICSPRDC
CICSPRDD
CICSPRDE
CICSPRDF
JOBWXYZ
JOBZYXW
========
CMD
COUNT
-----31,069
19,533
3,639
520
41
5,224
1,201
52
-----61,341
MEAN
MEAN
MEAN MEAN MEAN
ENQTIME DURATION
CMD PST TOT
SECS SECONDS
TIME ECB
IO
-------- -------- ------- ---- ---0.000021 0.00025 0.00005 3.15 0.01
0.000023 0.00021 0.00005 3.24 0.00
0.000022 0.00046 0.00005 3.79 0.03
0.000023 0.00089 0.00007 3.63 0.09
0.000016 0.00107 0.00013 4.68 0.12
0.000020 0.00055 0.00009 3.51 0.05
0.000033 0.00049 0.00005 4.00 0.02
0.000880 0.00022 0.00012 3.50 0.50
-------- -------- ------- ---- ---0.000022 0.00029 0.00005 3.27 0.02
61
B.THRUPUT ISSUES. SOLUTIONS. 15. USER WORKLOAD.
b. ONLINE
BY TP-TRANS-NAME
TP
MEAN
MEAN
MEAN MEAN MEAN
TRANS
CMD ENQTIME DURATION
CMD
PST
TOT
NAME COUNT
SECS SECONDS
TIME
ECB
IO
----- ------ -------- -------- ------- ----- ----NATP 14,411 0.000022 0.00024 0.00006 2.06 0.03
NATX
123 0.000024 0.00018 0.00007 1.41 0.02
COB1
963 0.000023 0.00045 0.00004 2.54 0.04
COB2
4,072 0.000023 0.00020 0.00009 2.39 0.00
: :
:
: :
: :
: :
:
:
62
B.THRUPUT ISSUES. SOLUTIONS. 15. USER WORKLOAD.
b. ONLINE
BY GLOBAL-USER
TERM OR
CMD
TRAN ID COUNT
------- ----CICSD2.K 1407
CIC92111
7
CICSD1$:
7
CICSD2.K
5
MEAN
MEAN
MEAN MEAN MEAN
ENQTIME DURATION
CMD
PST TOT
SECS
SECONDS
TIME
ECB
IO
--------- --------- -------- ----- ---0.000006
0.00002 0.00002 6.14 0.00
0.000011
0.00019 0.00005 3.43 0.00
0.000011
0.00018 0.00005 2.86 0.00
0.000012
0.00079 0.00008 2.60 0.20
63
B.THRUPUT ISSUES. SOLUTIONS. 15. USER WORKLOAD.
b. ONLINE
WHODID: EXTRACT
TIME
**
JOBNAME TP-TRANS-NAME
**
NAT-LOGON
NAT-PROGRAM
(HD='PROGRAM')
CID
(PF=H8)
CMD
FILE
(PF=N3)
ISN
(PF=N8)
RESPONSE-CODE (HD='RSP','CD ')
DURATION
TOTAL-IO
(PF=N2)
L3-SEQUENCE
(PF=A2)
COPT1 COPT2 ISNQ (PF=N4)
SUB-CODE
(HD='RSP','SUB')
WHERE
GLOBAL-USER = '
'
INSIGHT-LINES = 300;
64
B.THRUPUT ISSUES. SOLUTIONS. 15. USER WORKLOAD.
b. ONLINE
TIME
-----175059
175059
175059
175059
175059
175059
175059
175059
175059
175059
175059
175059
: :
PROGRAM
-------NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
NATPXTNB
: :
CID
-------07500301
07500301
07500301
07500301
07500301
07500301
07500301
07500301
07500301
07500301
07500301
07500301
:
:
C
C C
M FIL
RSP DURAT TOT L3 O O
D NUM
ISN CD
SECS IO SEQ 1 2
-- --- ------- --- ----- --- --- - L3 21 114,512
0 0.000
0 TS . V
RC
0
0
0 0.000
0 .. S I
L3 21 114,512
0 0.000
0 TS . V
RC
0
0
0 0.000
0 .. S I
L3 21 114,512
0 0.000
0 TS . V
RC
0
0
0 0.000
0 .. S I
L3 21 114,512
0 0.000
0 TS . V
RC
0
0
0 0.000
0 .. S I
L3 21 114,512
0 0.000
0 TS . V
RC
0
0
0 0.000
0 .. S I
L3 21 114,512
0 0.000
0 TS . V
RC
0
0
0 0.000
0 .. S I
:
:
:
65
B.THRUPUT ISSUES. SOLUTIONS.
16. NEW DATABASE VS. SMP.
WHAT SMP SOLVES:
WHAT SMP DOES NOT SOLVE:
THRUPUT
1. CPU
(10-15% INCREASE)
2. IO
(LOSS OF CACHING
200% INCREASE)
3. SVC DEPENDENCY
(CONVERSIONS)
66
B.THRUPUT ISSUES. SOLUTIONS.
16. NEW DATABASE VS. SMP.
POINTS TO CONSIDER BEFORE SMP:
- OTHER ADABAS TUNING (THIS DOC)
- CURRENT / PLANNED CPU UPGRADES
- MAX OF CURRENT ARCHITECTURE:
CMDS / SEC
vs. CMDS / SEC OF CPU
CMDS / SEC OF CPU = 1 / (CPU SECS / CMD)
( FOR CPU SECONDS PER COMMAND, SEE NEXT SECTION )
67
B.THRUPUT ISSUES. SOLUTIONS.
17. SEE CPU ISSUES.
18. SEE IO ISSUES.
68
B. CPU ISSUES. TOOLS.
1.DDPRINT FROM NUC SHUTDOWN.
2.DSTAT AND IOF.
3.ADAREP SYNS-60 CHECKPOINT.
4.CLOG COMMAND TIME.
5.STROBE,OMEGAMON,etc.
69
B. CPU ISSUES. TOOLS. 1.DDPRINT FROM NUC SHUTDOWN.
DURATION
WAIT-TIME
CPU-TIME
00296:30:07 HOURS = 1,067,407 SECONDS
00205:17:52 HOURS
00105:53:32 HOURS =
381,212 SECONDS
CPU = 35.7 %
A D A B A S
THREAD
1
2
:
TOTAL
EXECUTED
961341,857 CALLS
NUMBER
983037,373
869948,405
: : :
961341,857
CORRECTED= 3,333,625,439
0.00011 CPU SECS/CMD
70
B. CPU ISSUES. TOOLS. 1.DDPRINT FROM NUC SHUTDOWN.
DDPRINT HISTORY
DATE
20010715
20010727
20010805
20010827
DB
099
099
099
099
TOTAL CMDS
3045887222
2438702234
3440990001
3333625439
HOURS
286.6
215.3
307.9
296.5
TOTALIO
275519717
216822796
317241601
287554044
CM/S IO/CM CPU/CM
2952 0.09 .00012
3146 0.09 .00012
3104 0.09 .00012
3123 0.08 .00011
71
B. CPU ISSUES. TOOLS.
2.DSTAT AND IOF.
TSO ISPF IOF
JOBNAME--JOBID--SYID----CPU-----I/O----STEP-----PROCSTEP
DB099NUC S20260 PRD2 21:05:50 7281060 DB099NUC NUCSTEP
= 75,950 SECONDS
NUC SYSLOG OF DSTAT
ADAN17 0099 2001-09-11 12:38:59 NR. OF COMMANDS=595042096
= 595,042,096
0.00013 CPU SECS/CMD
72
B. CPU ISSUES. TOOLS. 3.ADAREP SYNS-60 CHECKPOINT.
ADARUN INTNAS=3600 or 900
ADAREP CPEXLIST
SYNS 60 2001-09-11 12:14:19 718 204371 DUAL ADABAS
NUCLEUS STATISTIC
NR OF CALLS =
12955548
NR OF PHYSICAL IO =
1422790
NR OF LOGICAL IO =
42672537
TIME =
2001-09-11 12:14:19
CPU USED =
00:28:33
= 1,713 SECONDS
= 12,955,548 CMDS
0.00013 CPU SECS/CMD
73
B. CPU ISSUES. TOOLS. 4.CLOG COMMAND TIME.
ALSO “ADABAS THREAD TIME”
DEFN. TIME IN THREAD MINUS WAITS: IO, OVERLAP
UNITS CONTROL BLOCK: 16 MICROSECONDS, 0.000016
C
MEAN
MEAN
MEAN
M
CMD TOT
DURAT
CMD
D
COUNT
IO
SECS
TIME
A1
1,558 0.24 0.00060 0.00026
BT
71 0.00 0.00012 0.00001
CL
1,759 0.00 0.00026 0.00000
ET
1,526 0.00 0.00016 0.00001
L1 62,993 0.00 0.00013 0.00005
L2
7,793 0.02 0.00026 0.00005
L3 171,553 0.86 0.00164 0.00011
L4
722 0.00 0.00035 0.00025
L6
450 0.15 0.00082 0.00025
L9
4,250 0.02 0.00021 0.00007
N1
304 3.05 0.00724 0.00105
RC 83,828 0.00 0.00010 0.00001
:
:
:
:
: :
:
:
== 385,841 0.42 0.00089 0.00008
C
MEAN
MEAN
MEAN
M
CMD TOT
DURAT
CMD
D
COUNT
IO
SECS
TIME
A1
1,759 0.00 0.00035 0.00025
BT
86 0.00 0.00013 0.00001
CL
1,095 0.00 0.00014 0.00000
ET
1,857 0.00 0.00015 0.00001
L1 82,262 0.00 0.00012 0.00005
L2 12,293 0.00 0.00016 0.00004
L3 121,410 0.00 0.00014 0.00006
L4
919 0.00 0.00037 0.00025
L6
541 0.00 0.00034 0.00024
L9
5,467 0.00 0.00018 0.00007
N1
88 0.00 0.00124 0.00105
RC 110,543 0.00 0.00009 0.00001
:
:
:
:
: :
:
:
74
== 385,119 0.00 0.00013 0.00005
B. CPU ISSUES. TOOLS. 4.CLOG COMMAND TIME.
CMDTIME
CPUSECS
700
600
500
400
300
200
83
79
75
77
75
73
71
69
67
65
63
61
59
57
55
53
51
49
47
45
43
41
39
37
35
33
31
29
27
25
23
21
19
17
15
13
11
9
7
5
3
1
0
81
100
B. CPU ISSUES. TOOLS. 5.STROBE,OMEGAMON,etc.
ADDRESSES OF HIGH CPU
OMEGAMON-CPU-NUC-CPU-ADARUN:
| LMOD Addr
|
000E10E0
|
00FCA000
|
000BC970
|
00011428
|
000ADC00
|
00000000
|
0004FC00
|
0003E2C0
|
|
|
|
|
|
|
|
|
LMOD Name
ADAMPM
IEANUC01
ADANC8
ADAIOS
ADANC7
*-UNKN-*
ADANC1
ADACSH
|% of Samples
|
71.60
|
20.63
|
2.39
|
1.44
|
1.24
|
1.17
|
.84
|
.66
|
Address |
Offset |% of Samples
|
000C21D0 | + 005860 |
.00 |
|
000C2820 | + 005EB0 |
.00 |
|
000C2E70 | + 006500 |
.00 |
000C34C0 | + 006B50 |
.00 |
ADANC8 |
|
000C3B10 | + 0071A0 |
.00 |
|
000C4160 | + 0077F0 |
92.72 |
|
000C47B0 | + 007E40 |
.00 |
+____________+__________+____________+
76
B. CPU. SOLUTIONS.
1. BAD Sx COMMANDS.
2. BAD NON-DE SEARCHES.
3. LOOPING PROGRAMS.
4. HIGH CPU USERS.
5. OVER REPORTING WITH CLOG TOOLS.
6. ADARUN LBP.
7. ADARUN LFP, CID’S.
8. ADARUN NC.
9. ADARUN NU,NT.
10.DVT CONSTRUCTION.
11.COMPRESSION/DECOMPRESSION.
12.BAD CACHE PARMS.
13.ADABAS RELEASE.
14.CPU HARDWARE.
77
B. CPU. SOLUTIONS.
1. BAD Sx COMMANDS.
SEE A.THRUPUT SOLUTIONS:
12. EXCESSIVE
2. BAD NON-DE SEARCHES.
SEE A.THRUPUT SOLUTIONS:
13. EXCESSIVE NON-DE SEARCHES.
3. LOOPING PROGRAMS.
SEE A.THRUPUT SOLUTIONS:
15. USER WORKLOAD IMPROVEMENTS.
4. HIGH CPU USERS.
SEE A.THRUPUT SOLUTIONS:
15. USER WORKLOAD IMPROVEMENTS.
BAD
SEARCHES.
5. OVER REPORTING WITH CLOG TOOLS.
SEE A.THRUPUT SOLUTIONS: 6. EXCESSIVE CMD LOG REPORTING.
78
B. CPU. SOLUTIONS. 6. ADARUN LBP.
! COST OF BUFFER FLUSHES !
LBP=100M
HEADERS = 100M / 2544 = 39K
THE RESULTS:
TEST
RUNS
---4
6
4
3
LAB TEST:
CYCLE NUC BETWEEN RUNS
FILL LBP WITH X BLOCKS
PGM: LOOP (10K):
A1 SAME FILE/ISN
NO ET
C1 (COPT=F)
BLKS/HDRS AVER.NUC
RANGE OF
LBP=
IN LBP
CPU (SECS) NUC CPU
---- --------- ----------- -------150M <20
19 SECS
18 TO 22
150M >20K
43 SECS
39 TO 46
200M >20K
40 SECS
40 TO 41
200M >36.7K
62 SECS
61 TO 63
79
B. CPU. SOLUTIONS. 6. ADARUN LBP. BUFFER FLUSHES.
MEASURE:
1. SHUTDOWN DDPRINT or SYNS-5B
DURATION
00166:30:07 HOURS
=FLUSH EVERY 6.8 SECS
88,355 BUFFERFLUSHES WERE DONE
2. AOS . . . SYSTEM STATUS
Physical
Reads
Writes
ASSO
5341459
844025
DATA
60343074
274962
WORK
293117
4317457
PLOG
1905450
Logical Reads .......
0
Buffer Efficiency ...
38.0
Call Distribution
Remote Logical ......... 56620167
Remote Physical ........
0
Local Logical ......... 573248391
Local Physical ........
0
Logical Reads (binary) . 7FFFFFFF
No. of HQEs active .....
7
No. of UQEs in User Queue.
296
Format Translations .44825232 No. of CQEs waiting in CQ .
0
Format Overwrites ...
3818
Total intern. Autorestarts . 8230
Throw Backs for ISN .
344 No. of PLOG switches .......
39
Throw Backs for Space.
0 No. of Bufferflushes ...... 801279
B. CPU. SOLUTIONS. 6. ADARUN LBP. BUFFER FLUSHES.
MEASURE:
1. APAS INSIGHT.
MIN
DATE
010910
010910
010910
010910
MIN
TIME
154500
160000
161500
163000
BUFFER: SUMMARIZE
MIN(DATE) MIN(TIME)
COUNT
SUM(TOTAL-IO)
BUFFER-FLUSHES-INTERVAL
BUFFER-FLUSHES-SESSION
INTERVAL 15 MINUTES
INSIGHT-LINES=300;
CMD
COUNT
4,701,761
4,663,621
3,320,363
3,506,226
SUM
TOT
IO
199,258
242,943
256,838
223,657
BUFF
FLUSH
INTER
8
10
10
12
BUFF
FLUSH
SESS
754
764
774
786
81
B. CPU. SOLUTIONS. 6. ADARUN LBP. BUFFER FLUSHES.
a. MINIMIZE BUFFER FLUSHES:
ADARUN LP=65,530
? LARGER WORK BLKSIZE ?
ADARUN LFIOP=
< ¼ LBP
>AVER. WRITE IO’S
EG. 4M TO 5M
b. MONITOR LBP SIZE:
ADARUN LBP=40-60M
USE ADABAS CACHING
82
B. CPU. SOLUTIONS. 7. ADARUN LFP, CID’S.
PROGRAM
FB=AA,020,A,..
TRANSLATE
ADABAS ADARUN LFP=
IFB:
60B HEADER
24B ELEMENTS
TYPES OF IFBs
1. USER
: UID (4b) + CID (4b)
2. GLOBAL : ADDS5 (8b), HIGH-ORDER = X’11’
83
B. CPU. SOLUTIONS. 7. ADARUN LFP, CID’S.
ADABAS 7.1.2
1. FREE SPACE IS CHAINED.
2. GLOBAL IFBs ARE CHAINED (LIFO).
3. USER IFBs ARE CHAINED (OFF U.Q.E.).
4. U.Q.E. CONTAINS ADDRESSES OF LAST 5 IFBs.
5. SCAN FOR IFB TO OVERWRITE (FB OVERWRITES): ???
84
B. CPU. SOLUTIONS. 7. ADARUN LFP, CID’S.
SCAN FOR
EXISTING
(5 MOST RECENT)
(UQE CHAIN)
(G.IFB CHAIN)
TRANSLATE
TIME
SCAN FOR SPACE
SCAN FOR F.OVWR
85
B. CPU. SOLUTIONS. 7. ADARUN LFP, CID’S.
CM
L1
L1
L1
L1
L1
L1
TRANSLATE TIME PER X FLDS
NO.
CMD
MEAN
FLDS
COUNT CMDTIME
FBL
1
1,000 0.00006
8
10
1,000 0.00013
80
20
1,000 0.00018
160
50
1,000 0.00036
400
100
1,000 0.00066
800
200
1,000 0.00129 1600
TIME TO SCAN X IFBs
CM
L1
L1
L1
L1
L1
L1
L1
L1
L1
IFBS
/LFP
1
100
500
1000
2000
4000
8000
16000
20000
CMD
COUNT
1,000
1,000
1,000
1,000
1,000
1,000
1,000
1,000
1,000
MEAN
CMDTIME
0.00006
0.00007
0.00016
0.00026
0.00047
0.00088
0.00168
0.00362
0.00460
86
B. CPU. SOLUTIONS. 7. ADARUN LFP, CID’S.
a. IF NO G.IFBs MAX LFP -HIGHWATER < 99%
-NO FORMAT OVERWRITES
b. LIMIT G.IFB CHAIN: MOST USED
? ? ?
c. IF LARGE G.IFB CHAIN
TOO SMALL < LFP < TO BIG
WATCH CPU
DO NOT WATCH FB.TRANS
DO NOT WATCH FB.OVWR
87
B. CPU. SOLUTIONS.
8. ADARUN NC.
(1.5 x HIGH.WATER)
<
NC
<
(2 x HIGH.WATER)
<
NU
NT
< (2 x HIGH.WATER)
<= 99% OF COMMANDS
9. ADARUN NU,NT.
(1.5 x HIGH.WATER)
88
B. CPU. SOLUTIONS. 10.DVT CONSTRUCTION.
ELIMINATE BAD DESCRIPTORS.
FILE CANDIDATES:
LARGEST USED NI
LARGEST ASSO-IO
MOST UPDATE CMDS
L3DE: SUMMARIZE(SUBTOT=OFF)
COUNT
(PF=N9)
MIN(NAT-PROGRAM) (PF=A8)
MAX(NAT-PROGRAM) (PF=A8)
MIN(CID)
(PF=H8)
MAX(CID)
(PF=H8)
BY FILE
(PF=N3)
BY L3-SEQUENCE
(PF=A2)
WHERE
(FILE = 48 OR = 139)
AND (CMD = 'L3' OR = 'L6')
INTERVAL X HOURS
PRINT TO OUTFILE
TITLE='L3 DE USE';
89
B. CPU. SOLUTIONS. 10.DVT CONSTRUCTION. KILL BAD DEs.
SXDE: SUMMARIZE(SUBTOT=OFF)
COUNT
MIN(NAT-PROGRAM)
MAX(NAT-PROGRAM)
MIN(CID)
(PF=H8)
MAX(CID)
(PF=H8)
BY FILE
(PF=N3)
BY DESCRIPTOR
(PF=A2)
WHERE
(FILE = 48 OR = 139)
AND (CMD = 'S1' OR = 'S2'
OR = 'S4' OR = 'L9')
INTERVAL X HOURS
PRINT TO ETWAS
TITLE='SX DE USE';
LOGSB
90
B. CPU. SOLUTIONS. 10.DVT CONSTRUCTION. KILL BAD DEs.
RESULTS
Number of DE's:
Total DE bytes/recs:
Total No. of Records:
Used Data Storage Blocks:
Used Index Blocks:
BEFORE
-----2120
41,248
231.9M
10.3M
18.1M
AFTER
-----1951
34,671
232.6M
10.4M
13.4M
CHANGE
------8%
-16%
0%
+1%
-26%
Cmd time / update cmd:
Duration / update cmd:
.00110
.01221
.00087
.00696
+21%
+43%
91
B. CPU. SOLUTIONS. 11.COMPRESSION/DECOMPRESSION.
! PROGRAMS SHOULD SELECT ONLY THE FIELDS NEEDED !
! AVOID GENERALIZED VIEWS !
A. GENERAL : READS 214 FIELDS OFF FILE
B. SPECIFIC: READS 20 FIELDS OFF FILE
DO 100,000 READS WITH BOTH METHODS.
92
B. CPU. SOLUTIONS. 11.COMPRESSION/DECOMPRESSION.
RESULTS:
RECORD BUFFER
ELAPSED TIME
CPU OF USER
CPU OF ADABAS
SPECIFIC
-------115b
9.4 MIN
?
89 SEC
GENERAL
-------1718b
14.7 MIN
?
152 SEC
PERCENT
CHANGE
-----56%
90%
71%
ADABAS SQL: AVOID SELECT *
? FIELDS AT THE END VERSUS THE BEGINNING ?
? FIXED VERSUS COMPRESSED ?
93
B. CPU. SOLUTIONS.
ADARUN
12.BAD CACHE PARMS.
CDATAMAXS=1,000,000
CMAXCSPS=1
CDATADSP = 1 – 1,000,000
CACHE SIZE = 197 RABNS
94
B. CPU. SOLUTIONS.
13.ADABAS RELEASE.
14.CPU HARDWARE.
95
C. IO ISSUES. TOOLS.
1. IO / CMD
a. SYNS-60
b. CLOG
c. DDPRINT
2. IO TIMES
a. CLOG
b. CACHE STATS
c. OMEGAMON,SMF, etc
3. BAD PACKS
a. DDPRINT
b. OMEGAMON,SMF, etc
4. BUSIEST FILES
96
C. IO ISSUES. TOOLS. 1. IO / CMD
a. SYNS-60
(SEE B. CPU ISSUES. TOOLS. 3. SYNS-60
b. CLOG
IOREP: SUMMARIZE(SUBTOT=NO)
COUNT
(PF=N5)
MEAN(TOTAL-IO) (PF=N1.2)
SUM(ASSO-IO)
(PF=N5)
SUM(DATA-IO)
(PF=N5)
SUM(WORK-IO)
(PF=N3)
*
BY JOBNAME BY FILE BY CMD
*
INTERVAL 15 MINUTES
*
WHERE ETWAS
*
PRINT TO OUTDD
INSIGHT-LINES=300;
97
C. IO ISSUES. TOOLS. 1. IO / CMD
b. CLOG
CMD
COUNT
-----51,944
MEAN
SUM
SUM SUM
TOT
ASSO
DATA WORK
IO
IO
IO
IO
---- ------ ------ ---0.18 2,732 6,392
0
NOTE: IF ADARUN LFIOP>0, THEN WRITE IO’S ARE NOT INCLUDED
NOTE: IO’S FROM ADABAS CACHE ARE NOT SUBTRACTED OUT
( ACTUAL IO WILL BE LESS )
98
C. IO ISSUES. TOOLS. 1. IO / CMD b. DDPRINT
ASSO
DATA
WORK
PLOG
CLOG
TOTAL
READS
34656619
248086098
4305
1522
0
282748544
WRITES
19657470
1791425
27968723
24974701
0
74392319
A D A B A S EXECUTED 1865226,653 CALLS
ADAI03 DDWORKR1
4305 READS 27968723
ADAI03 DDDATAR1 248086098 READS
1791425
ADAI01 DDLOG
0 WRITES
ADAI03 DDPLOGR1
762 READS 12830079
ADAI03 DDPLOGR2
762 READS 12144623
ADAI03 DDASSOR1 34656620 READS 19657471
WRITES
WRITES
WRITES
WRITES
WRITES
NOTE: CACHE IO SUBTRACTED OUT
NOTE: IF ADARUN LFIOP>0, THEN WRITE IO’S ARE NOT RABNs
99
C. IO ISSUES. TOOLS. 1. IO / CMD b. DDPRINT cont.
VOL-SER
ADAB59
ADAB60
ADAB61
: :
ADABDB
ADABDC
ADABDD
TOTAL
HIGH RABN
(ASSO
600828)
(ASSO
1201668)
(ASSO
1802508)
: :
:
:
(DATA
5507695)
(DATA
5758045)
(DATA
6008395)
COUNT
6108205
3185966
29549429
: :
13777324
1453608
1
419381673<= VOLSER COUNT
NOTE: CACHE IO NOT SUBTRACTED OUT
NOTE: IO COUNT IS READ AND WRITES
100
C. IO ISSUES. TOOLS. 1. IO / CMD b. DDPRINT cont.
+ ------------ ---ASSO---- ---DATA---- ---WORK--- +
+ CACHE WRITES 103,043,539 83,878,641
0 +
+
READ EXCPS
3,380,279 68,183,677
0 +
+ CACHE READS 99,521,064 15,669,051
0 +
+ TOTAL READS 102,901,343 83,852,728
0 +
+
EFFICIENCY
96.7
18.6
N/A +
+-------------------------------------------------+
NOTE:
ASSO.READS + ASSO.WRTS + DATA.READS + DATA.WRTS
+ CACHE.ASSO.READS + CACHE.DATA.READS
= VOLSER.COUNT
EXAMPLE
34656619 + 19657470 + 248086098
+ 99,521,064 + 15,669,051
= 419381673
+
1791425
(+/- COUPLA IO’S)
101
C. IO ISSUES. TOOLS.
2. IO TIMES
a. CLOG
IOTIME: SUMMARIZE(SUBTOT=NO)
COUNT
(PF=N5)
MEAN(DURATION) (PF=N1.5)
MEAN(CMD-TIME) (PF=N1.5)
BY CMD BY TOTAL-IO (PF=N1 HD='IO')
WHERE ECBS = 1
AND TOTAL-IO LE 1
AND CMD(1) = 'L'
* AND FILE = 111
C
MEAN
MEAN
M
CMD
DURAT
CMD
D IO COUNT
SECS
TIME
-- -- ------ ------- ------L3 0 4,889 0.00018 0.00005
L3 1
276 0.00150 0.00022
L9 0
216 0.00022 0.00012
L9 1
3 0.01303 0.00023
102
C. IO ISSUES. TOOLS.
2. IO TIMES
b. CACHE
+-------------------------------------------------------+
+ 00022 DATA SPACE, ASSO, RABNS 14,986,838 THRU 15,008,837
+-------------------------------------------------------+
+
ALLOCATED, LA=01:34:16
+
+-------------------------------------------------------+
+ 3,555,713 CACHE WRITES +
5,368 BLKS IN CACHE +
+
416,188 READ EXCPS
+ 200,007,680 CACHE SIZE
+
+ 3,120,907 CACHE READS
+
48,350 BLKS/CACHE
+
+ 3,537,095 TOTAL READS
+
88,016 RABN TAB SIZE +
+
88.2 DSP EFFICIENCY+
36,446,432 MAX DSP USED +
+
0.526364 MAX NIOT (SEC)+
1.221436 MAX EXCPT(SEC)+
+
0.000003 MIN NIOT (SEC)+
0.001940 MIN EXCPT(SEC)+
+
0.000077 AVE NIOT (SEC)+
0.028199 AVE EXCPT(SEC)+
+-------------------------------------------------------+
103
C. IO ISSUES. TOOLS.
2. IO TIMES
c. OMEGAMON,EPILOG,MAINVIEW,ASTEX,etc
+==========================
|Volume SMF Dev
I/O
|Serial id
#
LCU Rate
|------ ---- ---- --- ---|ADABC2 CNT1 085F 00E 6.2
DASD Device Activity ========
-------- Time in Milliseconds
Total = IOSQ+Pend+Conn+Disc
-------- ---- ---- ---- 1
0
0
1
0
104
C. IO ISSUES. TOOLS.
3. BAD PACKS
a. DDPRINT VOLSER COUNTS (SEE EARLIER)
b. OMEGAMON,EPILOG,MAINVIEW,ASTEX,etc
+-----------------------------------------------| STARTED TASK = DB999NUC
JES NUMBER = 15859
| FROM
00:14 ON 10/04/01 TO 00:14 ON 10/05/01
+-----------------------------------------------|WAIT_REASON_____________TIME_____%_|
|USING CPU
9:35 H 39.9|
|ECB WAIT (W/ STIMER)
3:28 H 14.4|
|DISK ADAB58 1033 ACT 1:28 H
6.1|
|STC ELAPSED TIME
23:59 H
+------------------------------------
105
C. IO ISSUES. TOOLS.
4. BUSIEST FILES.
BIGFILE: SUMMARIZE(SUBTOT=NO)
COUNT
(PF=N9)
SUM(ASSO-IO)
(PF=N9)
SUM(DATA-IO)
(PF=N9)
SUM(DURATION) (PF=N6.1)
SUM(CMD-TIME) (PF=N6.1)
BY FILE (PF=N3)
FILE
39
86
196
93
108
90
95
COUNT
120,421,719
156,184,181
69,001,326
41,769,213
84,563,613
92,297,406
43,576,891
ASSOIO
DATAIO DURATION CMDTIME
6,723,425 16,394,000 145,645.6 8,994.4
3,855,282 9,389,882 117,031.2 11,787.5
2,627,615 7,888,037 88,886.2 6,548.5
3,244,980 8,372,643 88,467.9 4,491.9
3,973,571 8,485,772 79,516.9 6,737.5
4,928,471 6,343,287 71,601.1 5,684.3
1,834,556 1,878,552 35,708.2 3,020.4
SORT: BY COUNT, BY DURATION, BY ASSO/DATA IO, BY CMDTIME
106
C. IO ISSUES. SOLUTIONS.
1. FORWARD INDEX COMPRESSION.
2. REMOVAL OF UNNEEDED DESCRIPTORS.
3. CHOICE OF DASD.
4. ADARUN ASYTVS=.
5. FILE PLACEMENT.
6. BLOCKSIZE.
7. REORDERs.
8. REUSE ISN/DS.
9. ASSO/DATA PADDING.
10.WORK 3 CACHE.
11.WORK 2 CACHE.
12.ADABAS CACHING.
107
C. IO ISSUES. SOLUTIONS.
1. FORWARD INDEX COMPRESSION.
SEE NEW WITH ADABAS 7
2. REMOVAL OF UNNEEDED DESCRIPTORS.
SEE B.CPU SOLUTIONS: 10.DVT CONSTRUCTION.
3. CHOICE OF DASD.
BEYOND THE SCOPE OF THIS PRESENTATION.
4. ADARUN ASYTVS=.
SEE NEW WITH ADABAS 7.
LEAVE AS DEFAULT (=YES)
108
C. IO ISSUES. SOLUTIONS.
5. FILE PLACEMENT.
SEE DDPRINT VOLSER COUNTS (MINUS CACHE)
SEE
SEE SYSTEM REPORTS (OMEGAMON,RMF,ETC)
WORST SEE BUSIEST FILE REPORTS
SEE ADAREP FOR WHICH FILES PER VOLSER
SOLUTION
MOVE THEM, SPREAD THEM
109
C. IO ISSUES. SOLUTIONS.
6. BLOCKSIZE.
ADVANTAGES
1. MAY IMPROVE PERFORMANCE ON READS.
2. ALLOW LARGER COMPRESSED RECORD SIZES.
3. FURTHER AVOIDANCE OF WORK WRAPAROUNDS.
DISADVANTAGES ? ? ?
PROCESS
ADAORD RESTRUCTUREDB, ADAFRM ASSO,DATA,etc,
ADADEF
ADAORD STORE
110
C. IO ISSUES. SOLUTIONS.
6. BLOCKSIZE. ADARUN PARM CHANGES
DEVICE = 8391
(everywhere)
LBP = 2 X OLD.LBP
WORK: EITHER DOUBLE WORKSIZE or
LP
= .42 X OLD.LP
LWKP2 = .42 X OLD.LWKP2
DUALPLD = 8391
DUALPLS = 60 X PLOG.CYLINDERS
CASSOMAXS = 2 X OLD.CASSOMAXS
CDATAMAXS = 2 X OLD.CDATAMAXS
111
C. IO ISSUES. SOLUTIONS.
7. REORDERs. : WHAT, WHEN ?
8. REUSE ISN/DS: USUALLY ON, UNLESS
9. ASSO/DATA PADDING: MIN. FOR BIG FILES
112
C. IO ISSUES. SOLUTIONS. 10.WORK 3 CACHE.
WHEN TO:
1. LOTSA WORK READS (?)
2. JUST TRY IT ( 1 %).
3. Sx ANALYSIS
ISN
QUANT
0
50
100
:
7,600
=====
CMD
COUNT
247,283
101
17
: :
3
247,467
SX: SUMMARIZE
COUNT
MEAN(ISNQ)
MEAN(WORK-IO)
MAX(WORK-IO)
BY ISNQ (PF=N5
WHERE CMD(1) =
(PF=N7)
(PF=N4.1)
(PF=N2.1)
(PF=N3)
BY 50)
'S'
MEAN MEAN MAX
ISN WORK WORK
QUANT
IO
IO
2.7 0.0
0
85.8 1.0
1
134.6 0.9
1
: : : :
:
7,635.0 3.0
3
3.0 0.0
3
ADARUN CWORK3FAC= %
ADARUN CWORKSTORAGE=DATASPACE
113
C. IO ISSUES. SOLUTIONS. 10.WORK 3 CACHE.
1.DETERMINE W3
BLOCKS
ADAREP:
DD- I DEV I NMBR OF I NMBR OF
NAMES I TYPE I CYLS
I BLOCKS
WORKR1 I 3390 I
500 I
67491
LP=50000
MINUS ADARUN:
MINUS ADARUN: LWKP2=14000
W3 SIZE =
2.WORST CASE
DDPRINT:
NO. OF LISTS TBI -POOL LI = 100000
3491
29080 ( 29 %)
ONE ELEMENT= (52 + 4 X NSISN)= 352b (NSISN=75)
DIVIDE TBI BY ELEMENT SIZE
NO. OF LISTS
82
ADARUN CWORK3FAC= 82/3491 = 3 % OR LESS
ASSUME 1 WORK BLOCK PER LIST
114
C. IO ISSUES. SOLUTIONS.
10.WORK 2 CACHE (ADARUN CWORK2FAC=):
a. LESS THAN 50%
b. TRY IT
MONITOR BOTH WORK 3 AND 2 CACHE SPACE VIA CACHE STATS
RE-EVALUATE CWORK3FAC / CWORK2FAC IF A CHANGE IN:
1.
2.
3.
4.
WORK BLOCKSIZE
ADARUN LWKP2=
WORKSIZE
(JUST CWORK3FAC)
ADARUN LP= (JUST CWORK3FAC)
115
C. IO ISSUES. SOLUTIONS. 12.ADABAS CACHING.
THE RESULTS:
TOTAL.CMDS
1,509,095,970
2,438,702,209
3,440,990,996
2,729,211,969
TOTAL.RDS
CACHE.RDS
297,266,628 123,654,424
198,852,925 82,473,849
285,295,358 120,069,839
313,548,268 130,706,933
SAVE
PCT%
29.4
29.3
29.6
29.4
CBE
4.3
6.0
2.4
6.8
PBE
3.0
4.2
1.6
4.7
SAVDUR
0.0016
0.0008
0.0007
0.0012
CBE: CURRENT ADABAS BUFFER EFFICIENCY
PBE: PROJECTED NON-CACHE BUFFER EFFICIENCY
SAVDUR: AVERAGE DURATION SAVED PER COMMAND (RESULT=0.00085
SAVDUR= SUM OF ALL (CACH.READ * (AVE.EXCP - AVE.NIOT))
DIVIDED BY TOTAL.CMDS
116
C. IO ISSUES. SOLUTIONS. 12.ADABAS CACHING. ADARUNS
ADARUNS:
CACHE
CMAXCSPS
CASSOMAXS
CDATAMAXS
CDISPSTAT
=
=
=
=
=
YES
1
* USELESS SINCE V6.2
100-200M * WATCH ZEROES!
100-200M * WATCH ZEROES!
NO
XA vs DSP vs EXT
NOTE ADABAS REGION: 4b PER RABN IN RANGE
Eg. CDATADSP=2000000-3000000
1,000,000 RABNS
4M IN ADABAS
117
C. IO ISSUES. SOLUTIONS. 12.ADABAS CACHING. TYPE OF ?
GLOBAL CACHING
1.CACHE RABN RANGES,
CASSODSP=from-to
2.CAN START ANYTIME
3.REVIEW IF FILE MOVES.
4.CACHE MAINTAINED FIFO.
5.CAN CACHE JUST UI, AC,
NI, FCBs, FDTs, DSST
versus
FILE CACHING
1.CACHE FILE, CLASS,
ASSO/DATA/BOTH
2.MUST START AT NUC INIT.
3.ADAJUSTS IF FILE MOVES.
4.CACHE MAINTAINED BY
CLASS BY FIFO
5.CACHE ALL FILE ASSO.
CANNOT CACHE FCB,FDT,
SEPARATE UI,AC,NI.
118
C. IO ISSUES. SOLUTIONS. 12.ADABAS CACHING.
PICK TOP FILES BY:
WHO / WHAT TO CACHE
FNR
COUNT
ASSOIO
DATAIO
39 120421719 6723425 16394000
86 156184181 3855282 9389882
196 69001326 2627615 7888037
93 41769213 3244980 8372643
:
: :
: :
:
:
CMDTIME
DURATION
DURATION CMDTM
ASSO OR DATA IO
145645.6 8994.4
COUNT
117031.2 11787.5
CRITICALNESS
88886.2 6548.5
88467.9
: :
4491.9
: :
1. CACHE FCB-FDT,DSST.
2. CACHE TOP 10 TO 30 FILES UI / AC.
3. CACHE TOP
5 TO 15 FILES NI.
( WATCH NI USED RABNS !)
4. CACHE TOP
1 TO 2 FILES DS WHERE USED RABNS < 2M
119
C. IO ISSUES. SOLUTIONS. 12.ADABAS CACHING. HOW TO
1. CREATE CACHE TABLE OF REQUESTS.
2. AT DATABASE START RUN CACHE JOB:
STEP 1 – ADAREP
STEP 2 – CACHE PROGRAM
STEP 3 – ISSUE CACHE COMMANDS
CACHE PROGRAM
1.READ TABLE.
2.READ ADAREP FOR FILE, TYPE,
RABN RANGES.
3.DETERMINE USED RABN RANGE.
4.ADJUST TO GROWTH.
5.BUILD CACHE OPERATOR COMMANDS.
CACHE TABLE
*FNR
FCB
DSST
39
39
39
93
93
93
93
86
86
49
49
: :
TYPE
GROWTH
AC
UI
NI
AC
UI
NI
DS
AC
UI
UI
AC
100%
100%
0%
100%
100%
0%
0%
100%
100%
100%
100%
120
C. IO ISSUES. SOLUTIONS. 12.ADABAS CACHING. HOW TO
Fnr/Type
FCB... ASSO
DSST.. ASSO
DSST.. ASSO
DSST.. ASSO
.14 AC ASSO
.14 NI ASSO
.14 UI ASSO
.21 AC ASSO
.21 NI ASSO
.21 UI ASSO
: : : : : :
FromRabn
______31
____1306
____8765
__291973
__345527
_2312178
____8595
__469025
__336926
___13004
: :
ToRabn
____1305
____1850
____8825
__292153
__346253
_2318177
____8674
__469242
__337925
___13033
Count of Blocksize
Rabns * RabnCnt
____1275
5273400
_____545
2254120
______61
252296
_____181
748616
_____727
3006872
____6000 24816000
______80
330880
_____218
901648
____1000
4136000
______30
124080
TOTAL ASSO RABNS REQUESTED TO BE CACHED: 478328
TOTAL DATA RABNS REQUESTED TO BE CACHED: 77987
XA MEMORY REQUIREMENTS FOR ASSO RABNINDEX ARE: 1913312
XA MEMORY REQUIREMENTS FOR DATA RABNINDEX ARE: 311948
LIST OF CACHE COMMANDS TO BE ISSUED:
/F DB044NUC,CASSODSP=31-1305 FCB
/F DB044NUC,CASSODSP=1306-1850 DSST
121
A D A B A S DYNAMIC CACHING SESSION S U M M A R Y
6.8 CURRENT ADABAS BUFFER EFFICIENCY
4.7 PROJECTED NON-CACHE BUFFER EFFICIENCY
C. IO
51 ACTIVE RABN RANGES
SOLUTIONS
51 RABN RANGES DEFINED
------------ -----ASSO---- -----DATA---- -----WORK---12. ADABAS
CACHE WRITES
125,209,233
29,452,553
0
READ EXCPS
7,639,638
15,928,509
0 CACHING
CACHE READS
117,208,374
13,498,559
0
TRACKING
TOTAL READS
124,848,012
29,427,068
0
EFFICIENCY
93.8
45.8
N/A
-----------------------------------------------------MAX CACHE
224,640K
195,320K
DDPRINT
00007 DATA SPACE, ASSO, RABNS 13,550 THRU 71,653
ALLOCATED, LA=01:33:18
9,196,498 CACHE WRITES +
9,387 BLKS IN CACHE
809,794 READ EXCPS
+
230,031,360 CACHE SIZE
8,383,500 CACHE READS
+
55,610 BLKS/CACHE
9,193,294 TOTAL READS
+
232,432 RABN TAB SIZE
91.1 DSP EFFICIENCY+
53,160,008 MAX DSP USED
0.395320 MAX NIOT (SEC)+
77.639942 MAX EXCPT(SEC)
0.000004 MIN NIOT (SEC)+
0.001938 MIN EXCPT(SEC)
122
0.000052 AVE NIOT (SEC)+
0.024895 AVE EXCPT(SEC)
C. IO ISSUES. SOLUTIONS. 12.ADABAS CACHING. TRACKING.
SESSION
DATE
20010520
20010617
20010630
20010715
20010727
20010805
20010827
20010909
...
HOURS
670.3
310.6
358.8
286.6
215.3
307.9
296.5
334.0
STRUC CACHREADS
108NI 14362597
108NI
5831633
108NI
6920049
108NI
6097142
108NI
4824751
108NI
6360703
108NI
6759524
108NI
8383500
READEXCPS
1572347
699255
751012
620154
565448
724404
695901
809794
AVE.NIOT
0.000050
0.000051
0.000048
0.000046
0.000054
0.000052
0.000051
0.000052
PCT
CACH
5.3
5.3
5.6
5.9
5.9
5.3
6.1
6.4
PCT
EFF BLOCKS CACHWRTS
90.1 55224 15939575
89.2 55751 6532991
90.2 56042 7672941
90.7 56274 6719169
89.5 56700 5391946
89.7 56730 7087559
90.6 57274 7458004
91.1 58104 9196498
AVE.EXCP
0.019529
0.019611
0.018708
0.023230
0.022807
0.020300
0.021086
0.024895
123
THE END
124