7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 1/44
© Copyright 2015Wellesley Information Services, Inc.
All rights reserved.
Advanced performance
optimization with SAP BW 7.3
and SAP BW 7.4
r. !"arne !ergComerit
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 2/44
2
In #his Session
• $et practical tips and techni%&es for maintaining and cleaning an
SA' !W system for optimal performance, incl&ding 'SA
optimi(ation, compression, maintaining statistical c&)es, and
controlling gro*th, red&cing log file si(es, removing #' temporary
storage, #' error logs, and temporary data)ase o)"ects.
• +ed&ce the si(e of an SA' !W system )y as m&ch as 0- )y taing
steps s&ch as removing 'SAs, aggregating, and optimi(ing
InfoC&)es, and implementing the ne* /SA architect&re
•
o* to clean )atch ta)les and red&ce the footprints &nneededdata.
• /earn ho* to tae advantage of ne* performance feat&res in !W
.3 and !W .4
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 3/44
3
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
• emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 4/44
4
9:plore the &se of line item dimensions for fields thatare fre%&ently conditioned in %&eries. #his model
change can yield faster %&eries.
!W .3 ;nonA8A< InfoC&)e esign /ine Item imensions
• /ine item dimensions are )asically
fields that are transaction oriented
• 7nce flagged as a line item dimension,
the field is act&ally stored in the factta)le and has no ta)le "oins
#his may res< in improvements to %&ery speeds for c&)es not
in !WA or A8A
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 5/44
5
!W .3 ;nonA8A< InfoC&)e esign = igh Cardinality >lags
•
ighCardinality flag for large InfoC&)es *ith more than 10 million ro*s
• At this company there *ere 11 InfoC&)es *ith a ratio of more than 30- of the
records in the dimensions vs. fact ta)le
• SA' recommends for Inde:ing and performance reasons to flag these as ?high
cardinality@ dimensions. o*ever, it has minor impact to smaller c&)es.
• In this e:ample, there *ere fo&r medi&m and large InfoC&)es that are not
follo*ing the )asic design g&idelines, and s&)se%&ently had slo* performance
any companies sho&ld redesign large InfoC&)es *ith highcardinality to
tae advantage of the standard performance enhancements availa)le.
R e a l E x a m p l e
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 6/44
6
!W .3 ;nonA8A< S7 esign and /ocs on /arge 7racle #a)les
In this e:ample, many of the verylarge S7s are not partitioned,
and several o)"ects have over 250
million records
Additionally, 101 S7 o)"ects
*ere flagged as )eing reporta)le.#his res<ed in System Is ;SIs<
)eing created d&ring activation.
Com)ined, these res<ed in
fre%&ent locs on the 7racle
data)ase and failed parallelactivation "o)s
'artition S7s. #he loc on very large S7s d&ring parallel loads are *ellno*n and SA' has iss&ed several notes on the topic6 B3445 D7S o)"ect6
Activation fails 9A/7CED and 434 D7racle deadlocs, 7+A000B0.D
R e a l E x a m p l e
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 7/447
• SA' 8etWeaver !W .3 has a ne*, step)ystep *i(ard that allo*s yo& to generate data
flo*s from flat files or e:isting data so&rces
•A great )enefit is that the *i(ards
*or against any Info'roviderF
i.e., yo& can &se the *i(ards to
create loads from S7s to S7s
or InfoC&)es
#his *i(ard red&ces the n&m)er or man&al steps needed to load data. It
also simplifies the development process and maes 9#/ *or m&ch easier.
#he !W .3 ata>lo* $eneration Wi(ard
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 8/448
ata)ase 'erformance ;nonA8A systems<
• ata)ase statistics are &sed )y the
data)ase optimi(er to ro&te %&eries.
7&tdated statistics leads to
performance degradation.
• 7&tdated inde:es can lead to very poor search performance in all
%&eries *here conditioning is &sed ;i.e., mandatory prompts<
• #he c&rrent sampling rates for this e:ample *ere too lo*, and
statistics sho&ld only )e r&n after ma"or data loads, and co&ld )esched&led *eely
>or many systems, data)ase statistics are o&tdated and may ca&se data)aseperformance to perform significantly poorer than other*ise *o&ld )e the case.
Sampling sho&ld often )e changed and process chains may )e resched&led.
R e a l E x a m p l e
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 9/449
•
#he 7/A' Cache is )y defa< 100 ! for local and 200 !for glo)al &se
• #he system at this company *as cons&ming no more than
0! on average
• #his means that most %&eries *ere ree:ec&ting the same
data ;good hit ratio of over G0-<
#he 7/A' emory Cache Si(e Htili(ation
R e a l E x a m p l e
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 10/4410
SA' A8A and !W .4
>or !W .4 on A8A, SA' has contin&ed to move more of the process intensive
f&nctions from the application to the ! server
#he )enefits of this approach are dramatically faster dataactivation, data transformations, and %&ery e:ec&tions
• #his taes advantage of the performance improvements of an inmemory !
• It also red&ces the need for data transfers )et*een application and ! server
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 11/4411
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
• emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 12/4412
'reSteps = Cleaning &p o&r !W System
• o& can save significant amo&nts of *or )y doing a
clean&p effort )efore yo& start yo&r A8A migration
or !W &pgrade pro"ect
• >or e:ample, an international company had a !W system *ith over
10 #!, *ith only 3B #! in the prod&ction )o: and the remaining
data on their 8ear/ine Storage ;8/S< sol&tion
• #his cleaned !W system saved them potentially millions of dollars
in hard*are and A8A licensing costs
It is not &n&s&al to red&ce a !W system
si(e )y 2030- d&ring a clean &p effort
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 13/4413
#he SA'J!WJ7HS9E99'I8$ #as /ist
• If yo& are on .0 S'32 of higher, yo& can generate an SA' !W o&seeeping tas
list and get a&tomated help in cleaning the system *ees )efore &pgrading it
o& first have to install the program from SA' 8ote 12G2 )efore yo& cangenerate the SA'J!WJ7HS9E99'I8$ tas list &sing tcode S#C01
. +eassign re%&ests *ritten into the incorrect 'SA partition
. Kerify ataSo&rce segments assignment to 'SA
G. eletes the entries no longer re%&ired in ta)le +SILW10. Clear all 7/A' Cache parameters
11. +epair InfoC&)e fact ta)le indices at ata ictionary level
12. +eorgani(e and delete )oomar Is M vie* Is
1. Checs !W metadata *ith IC
2. elete +S## traces
3. elete !W statistical data 4. elete Aggregate data via deactivation
5. 9ns&re partitioned ta)les are correctly
inde:ed for 'SA
B. 9ns&re re%&est consistencies in the 'SA
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 14/4414
A #ool to elp to igrate and Clean Hp
• SA' has created a cocpit to6 Clean &p the SA' !W system +ed&ce system si(e Cond&ct prechecs
;readiness checs< Si(e the system >ind s&)optimal code ;i.e.,
transformations< /oo at ta)le distri)&tions
and loads
#here are over 235 tests in this
tool #hese tools are thans to SA's arc !ernard
and his team at SA' /a)s Canada
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 15/4415
ore #ips to ae the ata)ase Smaller
•
Hse *riteoptimi(ed S7s as first level data stores. #hese caneasily )e offloaded o&t of main memory in A8A and save yo& money.
• Eeep yo&r 'ersistent Staging #a)les ;'SA< clean. !#W6 #he 'SA is often not
needed at all in !W .4.
• If yo& are on !W .3 Service 'ac and A8A *ith at least Service 'ac 5,
the *riteoptimi(ed S7s and 'SAs are flagged as ?early &nload@ from the
A8A memory. #his *ill help yo& eep the system smaller and re%&ire less
memory.
• o& can also flag other InfoC&)es, S7s, ta)les, and partitioned as ?not
active@. If yo& do so, they *ill only )e loaded into memory *hen act&ally
re%&ired.
#he si(ing program in SA' 8ote 13BGB taes these si(esavings settings into acco&nt *hen si(ing yo&r A8A system
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 16/44
16
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
• emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 17/44
17
emo6 7ptimal SA' !W on A8A performance
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 18/44
18
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
• emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 19/44
19
12 'reSteps = Cleaning &p o&r !W System
1. Clean the 'ersistent Staging Area ;'SA< for data already loaded to S7s.2. elete the Aggregates ;s&mmary ta)les<. #hey *ill not )e needed again.
3. Compress the 9 and > ta)les in all InfoC&)es. #his *ill mae InfoC&)es
m&ch smaller.
4. +emove data from the statistical c&)es ;they start *ith the technical
name of 0C#CJ:::<. #hese contain performance information for the !Wsystem r&nning on the relational data)ase. o& can do this &sing the
transaction +SS#A# or the program +SS#A#JA#AJ9/9#9 to help
yo&.
5. /oo at the log files, )oomars, and &n&sed !9: %&eries and templates
;transaction +SN9/9#9<.
B. +emove as m&ch as possi)le of the #' temporary storage, #' error
logs, and temporary data)ase o)"ects. elp and programs to do this
are fo&nd in SA' 8otes 113G3GB and 110B3G3.
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 20/44
20
12 'reSteps = Cleaning &p o&r !W System ;cont.<
. >or *riteoptimi(ed S7s that p&sh data to reporta)le
S7s ;/SA approach<, remove data in the *rite
optimi(ed S7s. It is already availa)le in higher level o)"ects.
. igrate old data to 8ear/ine Storage ;8/S< on a smallserver. #his *ill still provide access to the data for the fe* &sers *ho
infre%&ently need to see this old data. o& *ill also )e a)le to %&ery it
*hen !W is on A8A, )&t it does not need to )e inmemory.
G. +emove data in &n&sed S7s, InfoC&)es, and files &sed for staging inthe !W system. #his incl&des possi)le reorgani(ation of master data
te:t and attri)&tes &sing process type in +S'C.
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 21/44
21
12 'reSteps = Cleaning &p o&r !W System ;cont.<
10. o& may also *ant to clean &p )acgro&nd information stored in theta)le +S!A#CA#A. #his ta)le can get very )ig if not managed. o&
sho&ld also consider archiving any Iocs and clean the t+>C %&e&es.
All of this *ill red&ce the si(e of the A8A system and help yo& fit the
system ta)les on the master node.
11. In SA' 8ote 0B4, SA' provides some ideas on ho* to eep the
!asis ta)les from gro*ing too fast in the f&t&reF if yo& are on Service
'ac 23 on !W .0 or higher, yo& can also delete &n*anted master
data directly ;see SA' 8ote6 1304<.
12. >inally, yo& can &se the program +SCK9+JIJH8HS9 to delete
any &n&sed dimension entries in yo&r InfoC&)es to red&ce the overall
system si(e.
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 22/44
22
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
• emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 23/44
23
emo6 !W .4 'erformance onitoring
In this demo *e *ill e:plore the !W .4 on A8A !A Cocpit >eat&res
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 24/44
24
emo6 !W .4 'erformance onitoring
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 25/44
25
!W .3 'erformance and Cocpit Capa)ilities
•
!W .3 monitors and cocpit capa)ilities also incl&de6 onitor of data)ase &sage and o)"ect si(es ;i.e., InfoC&)es, S7s<
O&ery &sage statistics are more visi)le ;similar to +S+#, +S+K, +S##<
We can see more of the &se of SA' 8etWeaver !W Accelerator and si(es
onitor for the act&al &se of 7/A'PL Cache and hit ratios
o& can no* selectively delete internal statistics in +SS#A#W )ydate thro&gh the &pdated +SS#A#JA#AJ9/9#9 A!A' program
#here is also a L 9ditor for coding and synta: assistance
Sol&tion anager has )een
&pdated to tae advantage of
these ne* monitors.
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 26/44
26
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
• emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 27/44
27
Converting Info'roviders andPor ata >lo*s
• While not re%&ired, InfoC&)es can )e
optimi(ed f&rther for A8A performance• #his )asically means ?flattening@ the
data str&ct&res and removing the
dimensions in !W from the physical
layer ;they still loo as if they e:ists<
any refer to this optional step as a ?f&nctional migration@ and do this after the A8A
migration has )een completed, often as a separate initiative ;see SA' 8ote 14G4G<
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 28/44
28
A8A 7ptimi(ed !W .4 S7s and 'erformance Improvements
•
!W optimi(ed S7s are no* created )y ?defa<@ in A8A
• #his means that data activations are done m&ch faster at the A8A
data)ase layer• #he change log is ept in a calc&lation vie* res<ing in smaller S7s
A8A optimi(ed S7s are also availa)le for !W .3, )&t no* they are created )y
defa<, so do not convert S7s to A8Aoptimi(ed. >ast activation is availa)le for
all standard S7s *itho&t conversion to A8Aoptimi(ed.
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 29/44
29
Converting Info'roviders andPor ata >lo*s
• #o help yo&, the SA' igration Cocpitalso allo*s yo& to migrate yo&r data
flo*s from 3.: to ata #ransfer 'rocesses
;#'s< as &sed in versions .0 and higher
• If yo& convert the data flo*s yo& get
)etter a&tomated data pacage #'optimi(ation, *hich loads data faster into
A8A.
• o& can also sim&late the data flo* )efore yo& do the real
conversion. When doing so, data is loaded for )oth versions
;3.: and .:< of the dataflo*s and the res<s are stored incl&ster ta)les. #he data is then compared to verify that the
dataflo* after migration calc&lates the same data as it did
)efore migration
• Since the differences are displayed separately, yo& can
analy(e the res<s and changes in details
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 30/44
30
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
• emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 31/44
31
9W esign Ks. 9vol&tion
An organi(ation has t*o f&ndamentalchoices6
1. !&ild a ne* *ell architected 9W
2. 9volve the old 9W or reporting system
!oth sol&tions are feasi)le, )&t organi(ations
that selects an evol&tionary approach sho&ld
)e selfa*are and monitor &ndesira)le add
ons and Q*oraro&nds@.
>ail&re to )rea *ith the past can )e
detrimental to an 9Ws longterm s&ccessR
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 32/44
32
ata esign #he Hse of /ayered Scala)le Architect&re ;/SA<
#he /SA consists logically of6 Ac%&isition layer armoni(ationP%&ality layer 'ropagation layer !&siness transformation layer +eporting layer Kirt&ali(ation layer
Since SA' !W .3 S'3 *e have
had a set of 10 templates to help
)&ild a layered data architect&re for
largescale data *areho&sing
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 33/44
3333
BWERP
Germany
FLEXIBLE
REPORTI
NG
Europe
(excl. Germany)
Europe 2
Europe 3
USA
Amerca! 1
Amerca! 2
A!a
Germany
BUSINES
S
TRANS.
Europe
(excl.Germany)
Europe 2
Europe 3
USA
Amerca! 1
Amerca! 2
A!a
Germany
DATA
PROPAGATI
ON
Europe
(excl.Germany)
Europe 2
Europe 3
USA
Amerca! 1
Amerca! 2
A!a
DATA
ACQUISITI
ON
"a#a
Ac$u!#on
"a#a Source
%ran!&er 'ule
n&o Source
E' %a*le
Germany
CORPORATE
MEMORY
Europe(excl.Germany)
Europe 2
Europe 3
USA
Amerca! 1
Amerca! 2
A!a
Germany
DIMENSION
AL
REPORTING
Europe
(excl. Germany)
Europe 2
Europe 3
USA
Amerca! 1
Amerca! 2
A!a
semantic
partitions
B /SA /ayers
41 total
o)"ects
9:ample6 C&rrent /SA ata Architect&re in SA' !W
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 34/44
3434
9:ample6 Simplified /SA ata Architect&re
BWERP FLEXIBLE
REPORTING
BUSINES
S
TRANS.
Europe
Amerca!
A!a
DATA
PROPAGATI
ON
Europe
Amerca!
A!a
DATA
ACQUISITION
"a#a Source
%ran!&er 'ule
n&o Source
E' %a*le
CORPORAT
E MEMORY
Europe
Amerca!
A!a
DIMENSIONAL
REPORTING
Remove 5
semantic
partitions
Remove 3 LSA
layers
41 shrins
to G total
o)"ects
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 35/44
35
Conformed
Reportale
!S"
Write
"ptimized
!S"
9W Comple: /ayered Architect&res
• #his !'C on !W system *as e:periencing
s&)stantial load performance iss&es• Some of this *as d&e to &nderlying SA' !W
config&ration, *hile some *as d&e to the
technical config&ration of the data store
architect&re and data flo* inside SA' !W
ro+uc#on !!ue! nclu+e+,1) "epen+en# -o*! no# runnn
!e$uen#ally/ .e./ loa+ &rom
Summary cu*e #o S#an cu*e
! !ome#me! execu#e+ *e&ore#e !ummary cu*e +a#a ! loa+e+
an+ ac#a#e+/ re!ul#n n ero
recor+! n #e !#an cu*e.
2) on la#ency # 6 layer! o&
SA/ "S!/ an+ n&ou*e!
*e&ore con!ol+a#on proce!!e!
can *e execu#e+.
#$%&'!()S #$%&'!(*S #$%&'!*+#$%&'!(3S #$%&'!((S
#$%&'!,( #$%&'!(7 #$%&'!(4#$%&'!,* #$%&'!(+
%& S-mmar
C-e
/#$%&'C*30
BPC Sta1in1
C-e
/BPC'C*(0
Con2olidation
C-e
/"C'C"0
ECC .*
A2ia5Pacific
ECC .*
orth5America
ECC 4.7
&atin5America
R63 3.(i
E
ECC 4.7
AS$A
Per2i2tent Sta1in1 Area /PSA0
Con2olidation Proce22e28
1) learn
2) oa+
3) oren Excane
4) Elmna#on!5) p#ma#on!
R e a l E x a m p l e
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 36/44
36
Write
"ptimized
!S"
>i:es to Comple: 9W Architect&re
•
#he fi: to this system incl&ded removing the conformed S7layer, *ith !9: flags for data stores that are never reported on.
• Also, the !'C staging c&)e served
little practical p&rpose since the data is
already staged in the $/ S&mmary c&)e
and the logic can )e maintained in the
load from this c&)e directly to the
consolidation c&)e.
#$%&'!()S #$%&'!(*S #$%&'!*+#$%&'!(3S #$%&'!((S
%& S-mmar
C-e
/#$%&'C*30
Con2olidation
C-e
/"C'C"0
ECC .*
A2ia5Pacific
ECC .*
orth5America
ECC 4.7
&atin5America
R63 3.(i
E
ECC 4.7
AS$A
Per2i2tent Sta1in1 Area /PSA0
Con2olidation Proce22e28
1) learn
2) oa+
3) oren Excane4) Elmna#on!
5) p#ma#on!
on#erm *ene&#! nclu+e+
re+uce+ +a#a la#ency/ &a!#er
+a#a ac#a#on/ le!! +a#a
replca#on/ !maller !y!#em
*acup! a! ell a! !mpl&e+
!y!#em man#enance.
R e a l E x a m p l e
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 37/44
37
9W ata esign Classical Hse of <i'rovider ints in !W
• If a %&ery has restrictions on this characteristic, the 7/A' processor is
already checed to see *hich part of the c&)es can ret&rn data for the %&ery.
#he data manager can then completely ignore the remaining c&)es.
'ro)lem6 #o red&ce data vol&me in each InfoC&)e,data is partitioned )y #ime period.
A %&ery m&st no* search in all Info'roviders to find
the data. #his is very slo*.
Sol&tion6 We can add ?hints@ to g&ide the %&ery e:ec&tion. In the
++EH/#I'+7KI8# ta)le, yo& can specify one or several
characteristics for each <i'rovider, *hich are then &sed to
partition the <i'rovider into !asicC&)es.
An entr in RR9:&;$PR"<=$; onl ma>e2 2en2e if a few attri-te2 of thi2
characteri2tic /that i2? onl a few data 2lice20 are affected in the ma@orit of? or
the mo2t important? -erie2 /SAP ote28 ((3. See al2o8 )4++ and (()+(0.
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 38/44
38
!W .3 and higher Semantic 'artitioned 7)"ects ;S'7<
• When data stores and InfoC&)es are allo*ed to gro* over time, the data load
and %&ery performance s&ffers
• 8ormally o)"ects sho&ld )e physically partitioned *hen the n&m)ers of
records e:ceed 100 200 million
o*ever, this may )e different depending on the si(e of yo&r hard*are and
the type of data)ase yo& &se• In SA' 8etWeaver !W .3 *e get an option to create a Semantic 'artitioned
7)"ect ;S'7< thro&gh *i(ards
o& can partition )ased on fields s&ch as calendar year, region, co&ntry, etc.
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 39/44
39
ata esign Semantic 'artitioned 7)"ects ;cont.<
•
When an S'7 is created, a reference str&ct&re eeps trac of thepartitions. #he str&ct&re is placed in the <i'rovider for %&erying.
S'7 Wi(ards create all ata #ransfer 'rocesses ;#'<,
transformations, filters for each data store, and a process chain
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 40/44
40
What Well Cover
•
!W .3 'erformance !asics• o&seholding tass in SA' !W .3 and .4
• emo6 7ptimal SA' !W .4 on A8A performance
• 12step preA8A migration clean&p tass
•
emo6 !W .4 performance monitoring• !W 7ptimi(ation after A8A migration
• /SA vs. /SA 7ptimi(ation and S'7s
• Wrap&p
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 41/44
41
Where to >ind ore Information
•
Introd&ction to SA' A8A )y !"arne !erg and 'enny Silvia, SA''ress 3rd edition.
• http6PPscn.sap.comPdocsP7C350GB SC8 SA' 8etWeaver !&siness Wareho&se .4
• http6PPhelp.sap.comPn*Jplatform elp SA' !W .4 *e) site
• http6PP***.stechno.netPsapnotes.htmlTvie*UsapnoteMidU153GB SA' !&siness content release note for !W .4
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 42/44
42
Eey 'oints to #ae ome
• !W .4 is the first release to tae f&ll advantage of SA' A8A
• Some of the f&nctions in .4 are also availa)le to nonA8A
c&stomers
• #he ne* Composite'roviders and the 7pen 7S Kie* maes A8A
and !W tightly integrated and capa)le to s&pport 9Ws )etter• & sho&ld )reafrom the past and start designing *ith the ne* !W
.4 feat&res inmind
• #he ne* monitoring feat&res in the !W !A Cocpit and the A8A
systems maes it m&ch easier to see *hat is occ&rring from adata)ase level for the non)asis team.
• !efore yo& si(e yo&r system, clean it &p and save hard*are costs.
• All c&stomers sho&ld consider the !W move to A8A in 2015V
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 43/44
43
o&r #&rnV
o* to contact me6
r. !erg))ergcomerit.com
'lease remem)er to complete yo&r session eval&ation
7/24/2019 BI2015 Berg Advanced Performance Optimization With SAP BW v05
http://slidepdf.com/reader/full/bi2015-berg-advanced-performance-optimization-with-sap-bw-v05 44/44
isclaimer
SAP, R/3, mySAP, mySAP.com, SAP NetWeaver®
, Duet®
, Partner!"#e, an" ot$er SAP pro"ucts an" services mentione" $erein as %ell as t$eirrespective lo#os are tra"emar&s or re#istere" tra"emar&s o' SAP A( in (ermany an" in several ot$er countries all over t$e %orl". All ot$er pro"uct
an" service names mentione" are t$e tra"emar&s o' t$eir respective companies. Wellesley )n'ormation Services is neit$er o%ne" nor controlle" *y
SAP.