Ways to optimize DB2 Query - view

Question

I have used the below SQL statement to create a DB2 view. Can someone help me on how to improve this view performance as it runs slow sometimes.
The one thing I noted is that it runs slow the first time but then for subsequent runs, it picks up speed. I have analyzed with the visual explain but when it runs slow, I don't see any index advises. So not really sure how to optimize this.
-- Generate SQL
-- Version: V7R4M0 190621
-- Generated on: 02/09/21 03:08:31
-- Relational Database: BIGBLUE
-- Standards Option: Db2 for i
CREATE VIEW IESQAFILE.PSCMPORDVW
AS
WITH INPROGRESS AS
(
SELECT
DIODR#
, DIDISP
, DIUNIT
, DISTST
, DIAPRV
, DIETAD
, DITRLR AS TRAILER_ID
, DIDR1
, DIETAT
FROM
IESQAFILE.LOAD
WHERE
DIETAD <> 0
AND DIETAT <> '0000'
ORDER BY
1
)
, STOPGROUP AS
(
SELECT
SOORD STOPORDER
, COUNT(*) STOPSREMAIN
, MIN(SOSTP#) NEXTSTOP
, MAX(SOAPPR) APPTREQ
FROM
PAVT.STOP
INNER JOIN
INPROGRESS
ON
DIODR# = SOORD
WHERE
SOARDT = 0
GROUP BY
SOORD
ORDER BY
1
)
, STOPAPPTS AS
(
SELECT
SOORD APPTORDER
, SOCUST STOPCUST
, SOEDA ETADATE
, SOETA ETATIME
, SOADT1 EARLYDATE
, SOATM1 EARLYTIME
, SOADT2 LATEDATE
, SOATM2 LATETIME
, SOCTYC NEXTCITY
, SOSTP# APPTSTOP
, SOST NEXTSTATE
FROM
IESQAFILE.STOPOFF
INNER JOIN
STOPGROUP
ON
STOPORDER = SOORD
AND NEXTSTOP = SOSTP#
)
SELECT
ORDER_NUMBER
, SHIPPER_ID
, SHIPPER_NAME
, SHIPPER_ADDRESS_1
, SHIPPER_ADDRESS_2
, SHIPPER_CITY
, SHIPPER_ST
, SHIPPER_ZIP
, SHIPPER_ZIP_EXT
, LOAD_AT_ID
, LOAD_AT_NAME
, LOAD_AT_ADDRESS_1
, LOAD_AT_ADDRESS_2
, LOAD_AT_CITY
, LOAD_AT_ST
, LOAD_AT_ZIP
, LOAD_AT_ZIP_EXT
, LOAD_AT_LATITUDE
, LOAD_AT_LONGITUDE
, EARLY_PU_DATE_TIME
, LATE_PU_DATE_TIME
, EARLY_DELV_DATE_TIME
, EST_REVENUE
, ORDER_DIV
, CONSIGNEE_ID
, CONSIGNEE_NAME
, CONSIGNEE_ADDRESS_1
, CONSIGNEE_ADDRESS_2
, CONSIGNEE_CITY
, CONSIGNEE_ST
, CONSIGNEE_ZIP
, CONSIGNEE_ZIP_EXT
, CONSIGNEE_LATITUDE
, CONSIGNEE_LONGITUDE
, TRAILER_TYPE
, ORDER_MESSAGE
, ADDITIONAL_STOPS
, CMDTY_CODE
, CMDTY_DESCRIPTION
, ORDER_MILES
, ORDER_WGT
, ORIGIN_CITY_CODE
, ORIGIN_CITY
, ORIGIN_ST
, DEST_CITY_CODE
, DEST_CITY_NAME
, DEST_ST
, PICK_UP_AREA
, PLAN_INFO
, NUMBER_LDS
, NUMBER_DISP
, SHIP_DATE_TIME
, NEW_PICKUP_AREA
, EQUIPMENT_NUMBER
, APPT_REQ
, APPT_MADE
, PRE_T_SEQ
, PRE_T_AREA
, LOAD_DISPATCHED
, CUST_SERV_REP
, NEGOTIATIONS
,
(
CASE
WHEN UNUNIT IS NOT NULL
THEN UNUNIT
ELSE ' '
END
)
UNIT_DISPATCHED
,
(
CASE
WHEN UNSUPR IS NOT NULL
THEN UNSUPR
ELSE ' '
END
)
DRIVER_MGR_CODE
, COALESCE(SUPNAM, ' ') DRIVER_MGR_NAME
,
(
CASE
WHEN UNFMGR IS NOT NULL
THEN UNFMGR
ELSE ' '
END
)
FLEET_MGR_CODE
, COALESCE(FLTNAM,' ') FLEET_MGR_NAME
,
(
CASE
WHEN UNTRL1 IS NOT NULL
THEN UNTRL1
ELSE ' '
END
)
TRAILER_ID,
DIDISP DISPATCH_NUMBER,
(COALESCE(BCMCNEW, ' ')) FED_MC_ID,
DIUNIT DISPATCHED_UNIT,
CASE
WHEN UNETAD <> 0
AND UNETAT = ''
THEN SMIS.CVTDATETIM(CHAR(UNETAD),'0000', (
SELECT
SUBSTR(DATA_AREA_VALUE, 1109, 2) AS TIMEZONE
FROM
TABLE(QSYS2.DATA_AREA_INFO('COMPAN', '*LIBL'))
)
)
WHEN UNETAD <> 0
THEN SMIS.CVTDATETIM(CHAR(UNETAD),UNETAT, (
SELECT
SUBSTR(DATA_AREA_VALUE, 1109, 2) AS TIMEZONE
FROM
TABLE(QSYS2.DATA_AREA_INFO('COMPAN', '*LIBL'))
)
)
WHEN UNETAD = 0
THEN '0000-00-00T00:00:00-00:00'
END AS ETA_DATE_TIME,
NEXTSTOP , CASE
WHEN SOARDT <> 0
AND SOARTM = ''
THEN SMIS.CVTDATETIM(CHAR(SOARDT),'0000', (
SELECT
SUBSTR(DATA_AREA_VALUE, 1109, 2) AS TIMEZONE
FROM
TABLE(QSYS2.DATA_AREA_INFO('COMPAN', '*LIBL'))
)
)
WHEN SOARDT <> 0
THEN SMIS.CVTDATETIM(CHAR(SOARDT),SOARTM, (
SELECT
SUBSTR(DATA_AREA_VALUE, 1109, 2) AS TIMEZONE
FROM
TABLE(QSYS2.DATA_AREA_INFO('COMPAN', '*LIBL'))
)
)
WHEN SOARDT = 0
THEN '0000-00-00T00:00:00-00:00'
END AS STOP_ARRIVAL_DATE_TIME
, CASE
WHEN SOLUDT <> 0
AND SOLUTM = ''
THEN SMIS.CVTDATETIM(CHAR(SOLUDT),'0000', (
SELECT
SUBSTR(DATA_AREA_VALUE, 1109, 2) AS TIMEZONE
FROM
TABLE(QSYS2.DATA_AREA_INFO('COMPAN', '*LIBL'))
)
)
WHEN SOLUDT <> 0
THEN SMIS.CVTDATETIM(CHAR(SOLUDT),SOLUTM, (
SELECT
SUBSTR(DATA_AREA_VALUE, 1109, 2) AS TIMEZONE
FROM
TABLE(QSYS2.DATA_AREA_INFO('COMPAN', '*LIBL'))
)
)
WHEN SOLUDT = 0
THEN '0000-00-00T00:00:00-00:00'
END AS STOP_DEPART_DATE_TIME
, ORBAMT ORDER_INV_AMT
, ORARST AR_STATUS_FLAG
, DISTST SETTLEMENT_FLAG
, DIAPRV APPROVED_FOR_PAY
, BCCARR CARRIER_CODE
, BCNAME CARRIER_NAME
, BCADDR CARRIER_ADDRESS_1
, BCADR2 CARRIER_ADDRESS_2
, BCCITY CARRIER_CITY
, BCST CARRIER_ST
, BCZIP CARRIER_ZIP
FROM
INPROGRESS
INNER JOIN
IESQAFILE.PSMAINORVW A
ON
DIODR# = ORDER_NUMBER
AND DIDISP = NUMBER_DISP
AND
(
SUBSTR(ORDER_NUMBER, 1, 2) <> 'DH'
AND SUBSTR(ORDER_NUMBER, 1, 1) <> 'M'
)
LEFT OUTER JOIN
IESQAFILE.STOPOFF
ON
DIODR# = SOORD
AND SOSTP# = 90
LEFT OUTER JOIN
IESQAFILE.LMCARR
ON
DIUNIT = BCCARR
LEFT OUTER JOIN
IESQAFILE.MMILES
ON
MMORD# = DIODR#
AND MMRECTYPE = 'D'
AND MMDSP# = DIDISP
EXCEPTION JOIN
IESQAFILE.ORDBILL B
ON
B.ORODR# = DIODR#
AND B.ORSEQ = ' '
AND ORARST = '1'
LEFT OUTER JOIN
STOPGROUP
ON
STOPORDER = DIODR#
LEFT OUTER JOIN
STOPAPPTS
ON
APPTORDER = STOPORDER
AND APPTSTOP = NEXTSTOP
LEFT OUTER JOIN
IESQAFILE.UNITS
ON
UNUNIT = DIUNIT
AND UNORD# = ORDER_NUMBER
LEFT OUTER JOIN
IESQAFILE.SUPMAST
ON
SUPCDE = UNSUPR
LEFT OUTER JOIN
IESQAFILE.FLTMAST
ON
UNFMGR = FLTCDE
WHERE
DIETAD <> 0
AND DIETAT <> '0000'
RCDFMT PSCMPORDVW ;

Answer 1

Without seeing the Visual Explain (VE) data (and it's not really something easily sharable here) giving you a "magic wand" answer not likely.
I'm surprised you don't see any indexes advised. But even so, you should be able to compare the VE data between "fast" and "slow" runs to see where the differences are. Make sure you have "View"-->"Highlighting"-->"Expensive Icons"-->... turned on. Also check "Options"-->"Attribute Detail" and "Options"-->"Graph Detail"...
Always a good idea to be using the latest and greatest version (1.1.8.6) of Access Client Solutions (ACS) which includes the Run SQL Script & VE components.
A couple of things jump out, an Encoded Vector Index (EVI) that include aggregates(count,min,max) might help here...
select soord stoporder
,count(*) stopsremain
,min(sostp#) nextstop
,max(soappr) apptreq
from pavt.stop
And index (or two) with derived columns substr(order_number,....) might help here:
from inprogress
inner join iesqafile.psmainorvw a
on diodr# = order_number
and didisp = number_disp
and (substr(order_number, 1, 2) <> 'DH'
and substr(order_number, 1, 1) <> 'M')
Again VE is your friend, you'll have to dig into what's going on and where the time is being spent.
Break the various parts of your statement down and try to optimize individual components when possible. Hopefully, the gains will remain when you put it back together.
One thing to remember, a view is performance neutral. It neither helps nor hurts performance.
You don't give an example of how you access the view, but have you considered making use of user defined table functions (UDTF)? While the query engine attempts to push down selection criteria, a UDTF's parameters make it easy for you to explicitly include selection criteria. You might even benefit from encapsulating the statement in a UDTF and building the view over the UDTF. You can see an example (via Generate SQL Source) for the message_queue_info view & UDTF in QSYS2 that IBM provided.