You are on page 1of 78

1

<Insert Picture Here>


Tuning All Layers Of E-Business Suite Performance Toics
Isam Alyousfi, Senior Director
Lester Gutierrez, Senior Architect, EBS Applications Performance
!
The following is intene to outline our general prouct
irection! It is intene for information purposes only,
an may not "e incorporate into any contract! It is
not a commitment to eli#er any material, coe, or
functionality, an shoul not "e relie upon in ma$ing
purchasing ecisions!
The e#elopment, release, an timing of any features
or functionality escri"e for %racle&s proucts remains
at the sole iscretion of %racle!
"
#
O$%ecti&es
'
%#er#iew of Performance (ethoology for %ptimizing E)
"usiness suite
'
%utline "est practices for Tuning E)Business layers
Tuning the Applications Tier
Tuning the *oncurrent (anager
Tuning the Data"ase Tier
Tuning the Applications
'
(1) As Basic Layers * Pro+ucts
B
r
o
w
s
er

Lis
ten
er
mod_oc4j
mod_proxy
mod_plsql
mod_osso
Apache
JVM1
App1 App2
O4
J
oc4j_sec!rity
L"A#
Oracle $orms
%epository
O&"
%A "ata'ase
( d' Listener
&dM
JVM2
App) App2
onc!rrent
#rocessor*
Mana+er
,xt-
procs
OAM
,
(1) As Basic Layers * Pro+ucts
-
Application + S,L
(anagement
System -esource
(anagement
Space
(anagement
Bac$up + -eco#ery
(anagement
Storage
(anagement
Data"ase
(anagement
Proceural stmts!
Proceural
Statement
E.ecutor
S,L stmts!
S,L Statement E.ecutor
PL/S,L
Bloc$
PL/S,L engine
S,L engine
*P0
SGA
(emory
S,L1ET
2D3
%*42
2A5A
6orms
2DB*
%A6ramewor$
7(L
BI Pu"lisher
Buffer *ache
86
*oncurrency
Disco#erer
%ptimizer
Access Path
Ine.es
9::,:::; S,L statements
<:: ; Proucts
61D, GL, AP, A-, %(, S*, =
*oncurrent manager
(LS
>::,::: ; PLS,L pac$ages
4:,::: ; Ta"les
%2SP
%*I
S%A
BPEL
%ut of (emory
.
?
Performance /et0o+ology for Otimi1ing of E-$usiness suite
'
@ow to Approach a Performance Issue
'
8here is the time going A Get the right iagnostics
8hereA
8hatA
8hyA
12
>:
Ho3 to Aroac0 a Performance Issue
14 5E6I7E the pro"lem clearly
14 8ATHE( the right ata to analyze the issue
>! Ientify the (OOT 9A:SE of the pro"lem, possi"ly gather aitional ata
>! Search for a ;7O<7 SOL:TIO7 or wor$aroun that aresses the root
cause of the pro"lem
>! If it is a prouct issue, PASS O7 the right information to support/e# through
the regular channels
>! Try to ientify a TE/PO(A(= <O(;A(O:75 to alle#iate the issue while
you get a prouct fi.
11
>>
5efine t0e ro$lem
' Get a clear unerstaning of the performance issue
8hat
8here
8hen
8hy
@ow
1)
><
5efine t0e ro$lem
SeBuence Diagram
1!
>C
8at0er +ata
. Start with Best Practices D EnoteD >><>:4C!>F
. S,L Tuning
/ Trace files
/ S,LT output EnoteD <>G>HI!>F
/ Trace Analyzer EnoteD <<4<I:!>F
/ A8- -eport EnoteD I4H94<!>F
/ A8- S,L -eport EawrsBrpt!sBlF
/ >>g S,L (onitoring
/ S,L Tuning A#isor
. PL/S,L Tuning
/ Prouct logs
/ PL/S,L Profiler EnoteD H:H::G!>F
. -eports Tracing
/ noteD >>>C>>!>
. Data"ase Tuning
/ A8- -eport EnoteD I4H94<!>F
/ ADD( report EnoteD <G:9GG!>F
/ Automate Session @istory EAS@F -eport
/ LT%( output EnoteD CG<C9C!>F
. 6orms Tuning
. 6orms Tracing EnoteD CICG4H!>F
. 6-D Log EnoteD 44G>99!>F
/ Generic noteD 4CH9G<!>
. (iletier Tuning
/ 25( Logs
/ 25( Sizing EnoteD C9<HG>!>F
/ 2DB* Tuning EnoteD <IHH9H!>F
. %S
/ %S8atcher EnoteD C:>>CI!>F
1"
(oot 9ause Analysis
'
8here is the time goingA
'
8ho is consuming timeA
'
8hyA
1#
<0ere is t0e time going > 8et t0e rig0t +iagnostics
'
Ientify the components that o the processing
-efer to the architecture
' which tierEsF play a part A
' which components within a tier come into play A
' H:J of issues are DB relate
1'
1,
(oot 9ause Analysis? 90ec@ t0e $asics
' 5ersion le#els
' -eBuire *onfiguration
Init!ora D 1otes C?9::?!> E-><F + <>9<:G!> E>>iF
' Latest patch le#els
' -ecommene Patch lists
1ote K-ecommene Performance Patches for the %racle E)Business SuiteL, <44:4:!>
Also chec$ for prouct specific notes
' A#aila"le 6i.es
' Best Practices in Performance for EBS
httpD//"logs!oracle!com/ste#en*han/
1-
<0ere is t0e time going > 8et t0e rig0t +iagnostics
'
If non)ata"ase Eclient sieF processing occurs within the flow
that was trace, you may see wait e#ents li$e
S,LM1et (essageJ
T*P Soc$etJ
PLS,L Loc$ Timer
'
%ther useful ata in ientifying where time is spent
-aw sBl trace file
Prouct logs
App ser#er logs such as http logs
*lient process cpu consumption from %S Efor e.ample psF
1.
Tuning t0e Alication Tier
)2
Tuning t0e Alications Tier
'
6orms
To reuce loa on the ata"ase ser#er, ensure users are optimally
utilizing the professional 6orms interfaces!
A#oi Blin Bueries
Pro#ie selecti#e criteria in 6in winows an L%5s
(inimize networ$ traffic an form open times
A#oi opening an closing forms across transactions
*om"ine forms from multiple)proucts onto a single menu
Generate complete S,L trace an 6orms -untime Diagnostics
E6-DF to e"ug performance issues
)1
<>
<0ere is t0e time going > 8et t0e rig0t +iagnostics
'
*lient Process *P0 *onsumption
@elp )N %racle Applications
6orms an Data"ase process i + *P0 Time
))
Tuning t0e Alications Tier
%*42/25(
'
0se one 25( per < *P0s
'O>:: concurrent users per 25(
)!
Tuning t0e Alications Tier
%*42
'
-esponse Time/*P0 0sage
'
%ut%f(emoryErrors
)"
Tuning t0e Alications Tier
'
6irst, rule out any S,L issues!
1ote CGIG?I!> on how to ena"le S,L trace for %A 6ramewor$
applications
'
If there are no ata"ase)relate issues, then
you nee to analyze the 25(
'
TechniBues you can useD
Threa umps P Sen to %racle Support for analysis
G* logsD *onfigure 25( sizing
Profilers ) Sen to %racle Support for analysis
(esonse Time*9P: Issues
)#
Tuning t0e Alications Tier
'
G* tuning
Total a#aila"le memory is the most important factor affecting G*
performance!
-e#iew the freBuency of collections, especially maQor collections
Ei!e! 6ull G*F
Ena"le #er"ose G* to tune heap sizes "ase on the G* traffic
Start withD -Xms1024M and Xmx1024M
Setting )7ms an )7m. to the same #alue a#oi the nee to memory
allocation uring runtime
If full G*s are too freBuent, consier increasing 7ms an 7m.
)'
Tuning t0e Alications Tier
'*ommon causes of %ut%f(emoryError
(emory Lea$ P Some o"Qects not cleane up
'Strong reache references, connection o"Qects, etc!
(emory @emorrhage P
'Some $in of infinite loop
'-eaing large ata sets consuming #ast amounts of memory in a short
perio of time
'*onnection lea$s leaing to #ast o"Qect allocations
'S,L Statements/caching may lea to e.cess memory use in 2DB*
li"raries
Sizing Issues P @eap is unersize for e.pecte loa
),
Tuning t0e Alications Tier
'
Symptoms of memory lea$
@eap usage increases steaily, o#er a relati#e longer perio of
time Ee!g! ser#eral hours or aysF
8hen the amount of memory lea$ is large enough, you will start
seeing continuous 6ull G* in the G* log
Different users may encounter %%( on totally unrelate flows
Generate @eap Dump files when %ut%f(emoryError occurs
'@elp %racle Support ientify the root cause of the lea$s
)-
OA 63@ ? AB/ Process I5
Process ID of the 25( hanling your session
'
*lient Process *P0 *onsumption
' %A 683 P 25( Process ID
' The KA"outL lin$ is at the "ottom of each page
' -eBuires 61DRDIAG1%STI*S profile set to
).
<Insert Picture Here>
Tuning t0e 9oncurrent /anager
!2
Tuning t0e 9oncurrent /anager
To ma.imize throughput for Qo"s which spawn parallel wor$ers Ei!e! Auto
In#oice, PayrollF, consier reucing the sleep time of the *onflict -esolution
(anager E*-(F
' Default is 9:s, consier G or >: secons
Increase the cache size Enum"er of reBuests cacheF to at least twice the
num"er of target processes
If the *( an Data"ase are co)locate, use IP* or "eBueath connection
*onsier eicating certain concurrent managers to process either short or
long running programs to a#oi Bueue "ac$up
Purge the 61D ta"les on a regular "asis using the KPurge *oncurrent
-eBuest an (anager DataL program
!1
Tuning t0e 9oncurrent /anager
'
8or$loa (anagement
Trim the fat! -e#iew an eliminate concurrent Qo"s that are not
reBuire an/or are not "eing use "y your users!
A#oi ena"ling an e.cessi#e num"er of stanar or specialize
managers
A common guieline is "etween >)< target process per *P0, "ut
this nees to "e "alance with resources reBuire "y online
acti#ity
!)
Tuning t0e 9oncurrent /anager
'
8or$loa (anagement
0se specialization rules an wor$ shifts to "in specific Qo"s to
specific time winows
'@elps a#oi scheuling resource intensi#e "atch reBuests uring
pea$ acti#ity
-escheule some programs to run when the concurrent managers
ha#e e.cess capacity an
A more Bueues if all Bueues are running at ma.imum capacity
only if resources permit!
!!
Tuning t0e 9oncurrent /anager
'
Transaction (anagers ) T(s
0se for synchronous online processingEe.DIn#entory TransactionsF
Ensure enough T(s e.ist to ser#ice the reBuest loa
'Set the profile K*oncurrentD8ait for A#aila"le T(L to > EseconF!
Set the sleep time on the T(s to a high num"er Ee!g! >: minutesF
'A#ois constant polls to chec$ for shutown reBuests
!"
<Insert Picture Here>
Tuning t0e 9lient C 7et3or@
!#
Tuning t0e 9lient
(inimize the "rowser memory footprint
'(inimize tool"ars, "rowser themes, e.tensions, plug)ins, helper applications,
an phishing controls
'Do not minimize E)Business Suite components E"ut minimize other appsF
'If there is a choice, use %A6/@T(L Esmaller memory profileF
0pgraing -A( is generally more important than *P0
Always $eep menus small Ecustomize if neeeF
'(enus are "oth memory an networ$ intensi#e
Oracle E-Business Suite 11i and 12 PC Client Performance
'http://www.oracle.com/apps_enchmar!/html/white-papers-e-usiness.html
!'
Tuning t0e 7et3or@
-outing Issues
Lost Pac$ets ue to e#ice saturation
DB an (T must "e co)locate on a local pri#ate networ$
Traffic from (T to *lientD
*onsier using Local pro.y in local offices Eglo"al single
instanceF, Enote >GC<HI!>F
!,
Tuning t0e 5ata$ase
!-
Tuning t0e 5ata$ase Tier 9onfiguration
' Ensure man+atory init4ora arameters are set correctly
(%S Docs C?9::?!> + <>9<:G!> E->< + >>i respecti#elyF
0se A8- a#isories to optimize SGA/PGA sizing
' (eDuire+ Patc0es C 6iEes*<or@aroun+s for ;no3n Issues
-ecommene Performance Patches )(%S Doc <44:4:!>
-eBuire Patching P Interopera"ility 1otes
'>>g-> *ertifie with >>i an ->< P (%S Docs 4G<IHC!> + ICG<I9!>
'>>g-< *ertifie with >>i P (%S Doc HH>G:G!>
'>:g-< 0pta$e patch "unle HC9<9HC, (%S Doc I?>:4?!>
!.
Tuning t0e 5ata$ase Tier 9onfiguration
' :se @ey 5ata$ase features
Auto (emory (anagement
' E.ception P Linu. @ugePages P (%S Doc I44I9?!>
System (anage 0no
Auto Segment Space (anagement
PL/S,L 1ati#e *ompilation
' 9on&ert to t0e OAT/ Ta$lesace /o+el
*onsoliates S of Apps ta"lespaces from 4:: to O><
' (%S Doc <4HHGI!>
Best practicesD L(T, ASS(, groups o"Qects "y size/type/access pattern
(igration utility can "e use to migrate a schema at a time
0se A0T%ALL%*ATE e.tent sizes
"2
Tuning t0e 5ata$ase Tier 5is@ IO
' I*O Otimi1ation
*onfigure + #erify irect I% is use if using file system atafiles
' Impro#es performance/scala"ility "y simulating raw e#ices, a#oiing
file system cache traffic
Ensure your I% su"system can hanle your pea$ I% loa + I%PS
' -AID, S!A!(!E! configs with sufficient spinles
' (easure with %-I%1 tool or in >>g
"BM0_%,0O1%,_MA2A3,%-AL&B%A4,_&O P MO0 "oc 525672-1
-euce the I% wor$loa
' Tune e.pensi#e S,L
' Trim the wor$loa of unnecessary tas$s
' Do C7 less I% on a#erage with >>g A#ance *ompression %ption
"1
Tuning t0e 5ata$ase Tier Statistics 8at0ering
' 0se only 61DRSTATS or the Gather Schema / Ta"le Statistics *oncurrent
Programs
Do 1%T 0SE the analyze or "msRstats comman irectly! It is not supporte, an
can result in su")optimal plans
' 0se the GAT@E-RA0T% option to gather incrementally
' Analyze all schemas at >:J, then specific o"Qects at .J
Due to ata s$ew, some ta"les "enefit from higher sampling J
6or >>g, can use #alue of zero EA0T%RSA(PLERSITEF
' Disa"le the >:g/>>g automatic Qo" to gather stats
")
Tuning t0e 5ata$ase Tier Statistics 8at0ering
' *onsier loc$ing stats for #ery #olatile ta"les once a representati#e set
is gathere
Some Interface ta"les
%ther e.amplesD apRselecteRin#oices, wshRprRwor$ers
0se 61DRSTATS!L%ADR7*L0DRTAB to s$ip the ta"le
' Do not gather statistics e.cessi#ely on entire schemas or the entire
ata"ase such as nightly or wee$ly
' 61DRSTATS now supports Data"ase >>g E.tene Stats! Patch
?4I:>?9! Also, in patch set ??9<H>G
"!
Tuning t0e 5ata$ase Tier Performance /anagement
' Esta$lis0 $aselines for +ifferent 3or@loa+s
0se A8- "aselines + iff reports

DBMS_WORKLOAD_REPOSITORY or Enterprise (anager


'%S metrics with tools such as %S 8atcher P (%S Doc C:>>CI!>
Timings/Traces for critical flows + concurrent programs
' :se 11gFs SGL Performance Analy1er HSPAI C SGL Plan
/anagement features HSP/I
*hange management performance impact
Sta"ilize e.ecution plans as neee + Buic$ly restore KgooL plans while
analysis an permanent solutions are foun
""
Tuning t0e 5ata$ase Tier - A<(
' -egularly -e#iew Automatic 8or$loa -epository ata
' -e#iew Top G Time E#ents section to etermine utilization +
"ottlenec$s an re#iew rele#ant etaile sections for these
"#
Tuning t0e 5ata$ase Tier - A<(
' 6or latc0 relate+ 3aits
%ften ue to non)shara"le S,L, su")optimal S,L which performs full ta"le
or full ine. scans, ynamic o"Qect creation/remo#al, etc!
-e#iew the latch Statistics section to etermine the hot latches
Trace some waiter an holer sessions to etermine actual cause
' 6or I*O relate+ 3aits
-e#iew S,L sections "y Logical/Physical reas an the Segment Statistics
sections "y I/%
' 6or enDueue relate+ or $uffer $usy 3aits
-e#iew the following sectionsD enBueue, segment statistics "y "uffer "usy
waits, row loc$ + ITL waits
"'
Tuning t0e 5ata$ase Tier - A<(
' Lin@ A<( 3it0 ASH +ata
(ap sessions + S,L to main "ottlenec$s
,uery Acti#e Session @istory EAS@F ata
Ena"le tracing for a few of the sessions with hea#y waits etermine the actual
cause an S,L statements!
-e#iew the A#isory sections in A8- to fine tune SGA + PGA
*orrelate top S,L in A8- to "usiness flows
Prouce an re#iew S,L A8- reports for top S,Ls EawrsBrpt!sBlF
(onitor %S Stats Einclue in >>g A8-F
' *P0 an (emory 0tilization + Paging, I/% statistics + response times
",
Tuning t0e 5ata$ase Tier 5B 9onsole
"-
Tuning t0e 5ata$ase Tier 5B 9onsole
".
'
If $ottlenec@ is IO relate+444
"d file se#uential read$ d file scattered read$ lo% file s&nc$ free uffer waits$ etc...'
Tune Top S,L is always the first step
(a.imize (emory a#aila"ility, allocate generously to "uffer cache, re#iew
A8- a#isories + monitor swapping + paging
0se AS( or alternati#ely use the SA(E methoology for " files
I% Su")systemD
' -AID >: still preferre most often for high en reBuirements Ieal a#g!
response times of U >:ms!
*hec$ for e.cessi#e/reunant ine.ing
*onfigure async I%, use Buic$ I% technologies
Tuning t0e 5ata$ase Tier
9ommon correcti&e actions to maEimi1e scala$ility
#2
'
If t0e $ottlenec@ is concurrency relate+ "en#
($ latch($ uffer us&($ etc '
0se glo"al hash)partitione ine.es for hot leaf "loc$s ) Ientify #ia A8- VTop
logical I%s "y SegmentV
Some of these waits are commonly cause "y "a S,L e.ecution plans
i!e! Dlatch: cache buffer chains, %buffer busy waits, read by other session
(a$e sure to use ASS( an %AT( ta"lespace moel!
Increase I1IT-A1S to alle#iate ITL contention
Increase seBuence caches
8or$ with support/consulting to e#aluate if ta"le/ine. partitioning will help
If the program allows, more freBuent commits E"atch sizeF can help "y reucing the
*- Econsistent reaF wor$ reBuire
Tuning t0e 5ata$ase Tier
9ommon correcti&e actions to maEimi1e scala$ility
#1
' If t0e $ottlenec@ is (A9 relate+ 444 "%c
($ %loal($ etc...'
6irst o all the steps mentione pre#iously!!!
*hec$ interconnect configuration )) 0se 2um"o frames!
Do not use parallel Buery/ml across noes
' 0se instanceRgroups + parallelRinstanceRgroups settings
-e#iew Parallel *oncurrent Processing + Application Affinity config
' ><!>!C introuce option to efine noe affinity at the program le#el P
*oncurrent )N Program )N Define
Perform -A* focuse A8- analysis =
Tuning t0e 5ata$ase Tier
9ommon correcti&e actions to maEimi1e scala$ility
#)
Tuning t0e 5ata$ase Tier
9ommon correcti&e actions to maEimi1e scala$ility
' (A9 A<( Analysis
>!5erify S,L e.ecution is optimize P >::49 trace/S,L (onitoring
<!*ollect A8- ata from all instances
) awrgrpt.sql and awrgdrpt.sql "cluster-wide ase and diff reports'
>!5erify I%, *P0 utilization an Interconnect messaging performance are within
accepta"le limits
<!Drill own on cluster wait e#ents ; KS,L orer "y *luster 8aitsL
C!*orrelate S,L e.ecutions on all instances accessing/moifying top segments in
Segment Statistics "y
. Glo"al *ache Buffer Busy + Buffer Busy 8aits
. *-/*urrent Bloc$s -ecei#e
. Logical -eas

#!
' To get full &alue from tracingJ
-aw trace must "e
' A complete, non)truncate trace file
))) *+,P -./E S.0E .S /.,.1E* 1O 12234 B51ES)))
' 6rom "efore the user action "egan To the point the DB session ensMM
T$prof output (0ST ha#e
' -untime E.ecution Plans an -ow *ounts + -ow Source Stats
' -un Time E.ecution Plan W K-ow Source %perationL XW KE.plain PlanL
Best practices, etaile how)to steps + recommene tracing patches ocumente
' (%S Doc >><>:4C!>
)Also re#iew Appeni. B P Getting In)(emory partial runtime stats
Tuning t0e 5ata$ase Tier T;P(O6? 8oo+K Ba+ C :gly
#"
Tuning t0e 5ata$ase Tier Goo T3P-%6
call count cpu elapsed disk query current rows
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Parse 1 0.00 0.00 0 0 0 0
Execute 7674758 780.79 78.70 0 1 0 0
!etc" 7674758 6#8.78 6757.80 5987 15#8465# 0 7674758
------- -------- -------- ---------- ---------- ---------- ---------- ----------
total 15#49517 7009.58 7486.5 5987 15#8465## 0 7674758

%isses in li&rary cac"e durin' parse( 1


%isses in li&rary cac"e durin' execute( 1
)pti*i+er *ode( ,--./)01
Parsin' user id( 17# 2recursi3e dept"( 4

/ows /ow 1ource )peration


------- ---------------------------------------------------
7674758 1)/5 ,66/E6,5E 2cr715#8465# pr75987 pw70 ti*e7717477779 us4
7675089 8E15E9 -))P1 2cr715#8465# pr75987 pw70 ti*e76919#78560 us4
7675089 1)/5 :8;<:E 2cr71#146176 pr716809 pw70 ti*e7608#0804#5 us4
5587864# 5,=-E ,>>E11 =? ;89E@ /)0;9 ,/.5/@.>)61.65 2cr71#146176 pr716809 pw70 ti*e75577878#7 us4
1#6#90#87 ;89E@ /,86E 1>,8 ,/.5/@.>)61.81 2cr7#4#9707 pr7#7706 pw70 ti*e71#6416980 us42o&Aect id..
7675089 5,=-E ,>>E11 =? ;89E@ /)0;9 /,.>:15)%E/.5/@.-;8E1.,-- 2cr7#0700#56 pr76#178 pw70 ti*e75#9908#69 us4
7675089 ;89E@ :8;<:E 1>,8 /,.>:15)%E/.5/@.-;8E1.:1 2cr7#0567 pr780 pw70 ti*e71961554#5 us42o&Aect ..
##
Tuning t0e 5ata$ase Tier
' <0y ugra+e EBS to
Internal tests show >:)<:J impro#ement in EBS "atch + reporting wor$loas
Significant -A* Performance impro#ements
%racle internal&s prouction EBS system on >>g->Y
'>>g-< scheule for the summer
Lo3er cost 3it0 A+&ance+ comressionK $etter*automate+ +iagnosticsK
lo3er-ris@ ugra+e 3it0 SP/*(AT*SPA J
Best Practices/Tips white papers
'twpRupgraingR>:gRtoR>>gRwhatRtoRe.pectRfromRoptimizer!pf
'httpD//www!oracle!com/appsR"enchmar$
#'
11g SGL Tracing En0ancements
1ew tracing parameter PLA1RSTAT
' firstRe.ecution Z allRe.ecutions Z aapti#eM
*onnection pool pro"lem to get row source stats is sol#e
3ey tracing patchesM liste on (%S Doc >><>:4C!>
Turn trace on for Qust > or a list of sBlRi&s
8ith real)time row source stats monitoring from
#[sBlRplanRstatistics P no more waiting hours for a complete trace to
get the right trace ata
E.pane e#ent synta. D
- a!"# s"ss$%n s"! "&"n!s 's(_!#a)" *s(+a,-s./012&.s12,34s526-71ds(8 9
- s(_!#a)" 1a$!:!#6" 2 /as"; 4$nd:!#6" 2 /as";<ans!a!:n"&"# 2 /$#s!_"x")6!$%n 2
a_"x")6!$%nss2ada<!$&" "&" : 12 42 =2 12 2 13
#,
11g Automatic SGL /onitoring C Tuning
#-
11g Performance En0ancements
' Otimi1er imro&ements
SP( ES,L Plan (anagementFP Buic$ relief while permanent solution arri#es
Better ensity estimates for s$ewe ata
6aster, more accurate statistics with A0T%RSA(PLERSITE
E.tene statistics ) aress preicate correlation
Aapti#e cursor sharing P to aress "in pee$ing issues
' Some ot0ersJ
Inter#al partitioning, Aitional partitioning schemes
Te.t ine.ing performance enhancements
-A* scales "etter + A8- is more -A* aware
(any others!!
#.
11g A+&ance+ 9omression
' Data compression for all ata types
' Structure, 0nstructure, Bac$up, 1etwor$
' -euces storage consumption "y
< to 4 times
' Impro#es rea performance
' Enhances memory, "uffer cache
utilization
' *omplete application transparency
' Benefits i#erse application
wor$loas
Initially
:ncomresse+
Bloc@
9omresse+
Bloc@
Partially
9omresse+
Bloc@
9omresse+
Bloc@
Emty
Bloc@
'2
'
Oracle 5ata$ase 11g eEten+s ta$le comression for
OLTP +ata
Support for con6entional *,/ Operations ".7SE81$ +P*91E'
1ew algorithm significantly reuces write o#erhea
' Batche compression minimizes impact on %LTP transactions
1o impact on reas
' -eas ata irectly in its compresse format
' (any reas actually see impro#e performance ue to fewer I/%s
an enhance memory efficiency

11g A+&ance+ 9omression
'1
Leaf Bloc$s
N
a
n
c
y
,

R
O
W
I
D
N
a
n
c
y
,

R
O
W
I
D
N
a
n
c
y
,

R
O
W
I
D
N
i
c
o
l
e
,

R
O
W
I
D
N
o
r
m
,

R
O
W
I
D
P
a
b
l
o
,

R
O
W
I
D
P
a
u
l
a
,

R
O
W
I
D
P
a
u
l
a
,

R
O
W
I
D
P
e
t
e
r
,

R
O
W
I
D
L
u
i
s
,

R
O
W
I
D
M
a
r
k
,

R
O
W
I
D
M
a
r
y
,

R
O
W
I
D
M
i
k
e
,

R
O
W
I
D
M
i
k
e
,

R
O
W
I
D
K
a
r
l
,

R
O
W
I
D
K
a
t
h
y
,

R
O
W
I
D
K
i
m
,

R
O
W
I
D
L
a
n
c
e
,

R
O
W
I
D
P
i
e
r
r
e
,

R
O
W
I
D
R
a
c
h
e
l
,

R
O
W
I
D
R
a
j
i
v
,

R
O
W
I
D
R
a
o
u
l
,

R
O
W
I
D
P
h
i
l
,

R
O
W
I
D
F
K
a
r
NP
P
h
!"
"
r
#
a
m
#
t
#
u
Di Lu Rh
11g A+&ance+ 9omression In+eE 9omression
. *ompress multi)column ine.es to
reuce the storage o#erhea of
repeate #alues
. Stores more $eys per leaf "loc$
. Less I/%,"etter performance of
range scans
')
11g A+&ance+ 9omression - 9omression A+&isor
' Estimate Storage Sa#ings for your case
A#aila"le in >>g -elease < or =
A#aila"le on %T1 M for pre#ious releases
Shows proQecte compression ratio for uncompresse ta"les
-eports actual compression ratio for compresse ta"les E>>% %nlyF

) http://www.oracle.com/technolo%&/products/dataase/compression/compression-ad6isor.html
'!
'OracleFs Internal E-Business Suite Pro+uction system eploye A*% in \:?
A#erage o#erall storage sa#ings ) C.
'Ta"le compression 4.
'Ine. compression <.
'L%B compression <!C.
)!1TB of reali1e+ storage sa&ings primary, stan"y + test systems
'Aitional "enefits also accrue in De# clones an "ac$up
'Payroll, %rer)<)*ash, AP/A- "atch flows, Self)Ser#ice flows run without
regression, ,ueries in#ol#ing full ta"le scans show speeup
11g A+&ance+ 9omression 3it0 E-Business Suite
'"
' 8hite paperD 0pgrae to %racle >>g ) Performance Best Practices
http://www.oracle.com/apps_enchmar!/html/white-papers-e-usiness.html
' Special attention to
Testing of >>g or >:g wor$loas with -eal Application Testing
(anaging S,L E.ecution Plans an minimizing performance
egraation with>>g S,L Plan (anagement
11g :gra+e Best Practices for EBS
'#
Tuning t0e Alications
''
Tuning t0e Alications
'
(y%racleSuport note K-ecommene Performance
Patches for the %racle E)Business SuiteL, <44:4:!>
-ecommene performance patches for all the moules an
tech! stac$ components are consoliate in this note
'
Apply latest -elease 0pate Pac$s
' 6unctionality, fi.es, an upates for all -elease >< functional
applications and the technology stac$
' Se#eral prouct family -0Ps inclue are also a#aila"le
ini#iually
',
Tuning t0e Alications
'
Apply latest -elease 0pate Pac$s
-elease ><!>!CD Patch ?<C?:?:
' (%S DocD >:H:?IC!>
-elease ><!> D Patch IC:C:C:
' (%S DocD IG<9>?!>
-elease ><!: ) -0P9D Patch 9I<H:::
' (%S DocD I4CC9H!>
' -e#iew EBS Benchmar$ Pu"lications
' http://www.oracle.com/apps_enchmar!/html/white-papers-e-usiness.html
'-
Tuning t0e Alications
'
Performance Best Practices
<or@flo3
' 6or "ac$groun engines #ia *oncurrent (anager, set the \Process Stuc$&
parameter to \1o&
Start a separate "ac$groun engine to hanle stuc$ processes with a low
freBuency Pi!e! once or twice a ay
' 0se eferre acti#ities to impro#e online response times for flows such as
Scheuling, P% Document Appro#al, etc!
' 6reBuently purge runtime ata E*onc ProgramD 61D86P-F
' -un [61DRT%P/sBl/wffngen!sBl to translate the acti#ity function calls into
static calls
' Disa"le retention on 8or$flow ,ueues
DBMS_A>ADM?ALTER_>@E@EA(6"6"_nam":B+41;#"!"n!$%n_!$m":B0CD
' 6or high #olume "atch processing in -A* use ITE(RT]PE to 1oe affinity
'.
Tuning t0e Alications
' Performance Best Practices
Purge C Arc0i&e
' Implement a regular purge process incluing transient ata as
well as o"solete transactional ata
' Purge ol ata "efore performing upgraes
' (any ocumente stanar programs a#aila"le
O<9: purge programs in -><
' Purge Portal
Purge programs can "e configure, initiate an monitore
' Set the e.ecution freBuency as well #iew history of purge
programs
-><D programs tagge with the KPurgeL program type
,2
' (untime Performance Testing Tis
0se Automate, scripte tools
' EBS Test Starte 3its E8inrunner/,TPF
Bunle ,A "ase automate scripts for EBS testing ) Patch H4:HHH9
' %racle Applications Testing Suite EAccelerators for EBSF
8e" an 6orms "ase flows
*omplement with user participation tests an "atch loa tests with freBuent
an critical Qo"s
-eferences
http://lo%s.oracle.com/ste6enChan/2::;/1:/oats_es_certified.html
http://lo%s.oracle.com/ste6enChan/2::;/:</e6olutionar&_steps_automated_testin%_es.html
Tuning t0e Alications
,1
(eferences
,)
(eferences
' -><!> ocumentation roamap EI?:?4<!>F
' K%racle E)Business Suite -elease ><!> Info centerL EH:9G?C!>F
' Data"ase preparation guielines for -><!> upgrae EI9>GI:!>F
' Patching 6A,s E4G?>G9!>, <<G>9G!>F
' 0sing stage or share APPLRT%P an istri"ute AD EIC4:<G!>,
CH4<4H!>, <C949?!>F
' %A( KPatch 8izarL o#er#iew an 6A, E?I9>HH!>, ?I99HH!>F
' AD *omman Line %ptions for -elease ->< E>:IH?IC!>F
' EBS ><!>!C Data (oel *omparison -eport E><?:HH9!>F
' -ecommene Performance 6i.es E<44:4:!>F
' ->< 0pgrae Sizing + Best Practices EC??C9<!>F
,!
S://A(=
5OFs44
'-ea an follow official ocumentation an
ha#e a proQect plan
'Engage Support early on to #aliate your
upgrae plan
'Ientify an e.ecute tas$s you can o toay
'Test X Test X Test X
'3eep trac$ of patches applie in test upgraes
'Be smart a"out using the right tools an
e.plore pu"lishe owntime reuction
techniBues
'%ptimize patch runs to suite your @/8
5O7FTs44
'Ignore incluing rele#ant functional S(Es in
the planning process from the "eginning
'S$ip faile or long running Qo"s as a regular
practice
'Drop ine.es
'-un apatch in serial moe
'Perform wor$arouns without appro#al from
support/DE5
'Insufficient time an/or S of rouns of upgrae
an critical functional flow testing
,"
Oracle Pro+ucts A&aila$le Online
%racle Store
Buy %racle license an support online
toay at
oracle!com/store
,#
,'
,,
A
Q
&
,-

You might also like