Hi Perf Analytics in SAP NetWeaver 2004s

Embed Size (px)

Citation preview

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    1/64

    Session ID: IM205High Per form anc eAnaly t ic s in SAP

    Net Weaver 2004s

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    2/64

    Klaus Nagel, SAP AG

    Alexander Peter, SAP AG

    Ron Silberstein, SAP Labs, LLC

    Marc Bernard, SAP Labs, LLC

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    3/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 3

    Learn ing Objec t ives

    As a resul t o f t h is w ork shop, you w i l l be ab le t o : Explain the significant performance improvements delivered

    with the High Performance Analytics capabilities deliveredwith SAP NetWeaver 2004s

    Understand that HPA also reduces the maintenancesubstantially

    Specify appropriate scenarios for HPA and their easy

    application on top of existing data models and processes Understand SAPs strong commitment to investing in

    performance innovations, simultaneously keeping TCO at aminimum

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    4/64

    Overview

    HPA Configuration

    HPA Administration

    HPA in Detail

    Summary

    Appendix: Some Performance Features 2004s

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    5/64 SAP AG 2005, SAP TechEd 05 / IM205 / 5

    Ent erpr ise Business Int e l l igenc e: Chal lenges

    Enterpr ise

    Business

    In te l l igence

    RapidlyGrowing

    Data Volumes

    Requirements:

    Speed andFlexibility

    IncreasingNumber of

    Users

    Administration challenges Potential performance issues

    Increasing demand for more,better, and actionableinformation

    Natural data growth,RFID, POS, EDW,retention policies,

    Right-time information,fast query execution,ad hoc queries,

    More application areas,more casual users

    Challe

    nges

    forres

    ourc

    e

    allo

    catio

    n:ext

    ensive

    proa

    ctiv

    e

    maint

    enanc

    eeffort

    require

    d

    Challenges for resourceallocation: updating capacity

    planning may result inadditional hardware investment

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    6/64 SAP AG 2005, SAP TechEd 05 / IM205 / 6

    Som e Ways t o Im prove Query Per form anc e

    Per formanc e Feat ures

    HighPerformanceAnalytics

    Aggregates OLAP

    Cache

    ReportingAgent /

    InformationBroadcasting

    Summarized (or subset)

    materialized views ofInfoCube data

    Accelerates response for

    similar queries by cachingquery result sets, andreading from cacheinstead of DB

    Run popular queries in background

    (off hours) and push summaryviews of updated data to users Information Broadcasting

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    7/64 SAP AG 2005, SAP TechEd 05 / IM205 / 7

    Trad i t iona l Aggregat es Back ground / Contex t

    Genera l ly , aggregat es ... Are BI objects designed to improve query response times

    Structurally look the same as an InfoCube

    Are subsets of fact table data

    (Typically) store summary data in new, smaller, child InfoCubes

    Improve query response times by reducing the volume of data read

    Are transparent to users

    If an aggregate is available for a request, the system uses it automatically

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    8/64 SAP AG 2005, SAP TechEd 05 / IM205 / 8

    How Many Aggregat es Do You Need?

    You have to dec ide w h ic haggregat es you need

    To ensure good performance

    Pre-calculate aggregates for main queriesand reports

    Restrict scope and frequency of reporting

    This is a trade-off between flexibility andperformance

    You cannot create aggregatesfor all possible queries

    Thus, the typical strategy is to buildsomewhat general aggregates that morequeries can utilize, but the performancebenefit per query is less significant

    # characteristics

    #aggreg

    ates

    Aggregate Combinations

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    9/64 SAP AG 2005, SAP TechEd 05 / IM205 / 9

    A Searc h Engine for Analy t ic s?

    A new approac h The latest computer architectures make more intensive use of

    increasingly inexpensive memory resources

    Search engines deliver fast responses by giving read accesspriority over index updates

    But can an engine first conceived for text search be used toimplement fast search for structured data in tables?

    Yes SAP adapted its TREX search engine to implement

    A c apab i l i t y fo r h igh per fo rmanc e ana ly t ic s !

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    10/64 SAP AG 2005, SAP TechEd 05 / IM205 / 10

    HPA Capabi l i t y for SAP Net Weaver BI

    High Per form anc e Analy t ic s (HPA)A new transparent approach based on TREXto boost BI query performance

    Performance speedup factor between 10 and 100Without changing the BI user experience (transparent to users)

    HPA Engine

    Storage Query ProcessingDBMS TREXSAP NetWeaver Business Intelligence

    Database

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    11/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 11

    Termino logy

    TREX and HPA TREX : Search and classification engine, mainly for unstructured data

    HPA is built using TREX technology (indexing, retrieving, compression...) to handle structured data

    Note: HPA and TREX are two different installations

    The HPA engine cannot be used for standard KMC functionality

    A TREX installation cannot be used for high performance analytics

    Inside info: technically there is a lot of commonality but TREX-HPA role

    switching is not supported

    Index

    HPA index: Structure containing replicated data and indexes (one index

    per table) that represents the BI star schema for the HPA engine Indexing: Transfer of InfoCube data to the HPA server and processing

    and compressing it into the HPA index

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    12/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 12

    HPA Arc h i tec t ure

    SAP NetWeaver BIHigh Performance

    Analytics Capability

    Data Acquisition

    InfoCubes

    BI AnalyticEngine

    BusinessExplorer

    Any Tool

    AnySource

    Query &Response

    Indexing

    HPA engine responds to queries:joins and aggregates in run time

    HPA indexes loadedinto memory

    HPA creates and storesindexes for InfoCubes

    Q E t i

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    13/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 13

    Query Ex ec ut ion

    Query ExecutionWithout HPA

    Pre-calculation /Information Broadcasting

    Query ExecutionWith HPA

    Pre-calculation /Information Broadcasting

    OLAP Cache

    HPA

    1

    OLAP Cache

    Aggregate

    InfoProvider

    2

    3

    4

    The BI system checks the listed repositories (in theorder shown) to obtain the query result set

    HPA R t i l

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    14/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 14

    HPA Rat ionale

    High Performance Analytics

    Very fast queryresponse time

    Stable queryresponse time

    High scalability Low maintenance

    Performance

    improvementsby factor 10 100

    Independent of

    DB optimizer,aggregates, ...

    No aggregatemaintenance, minimized

    roll-up/change run

    Planned support of newhardware paradigms like

    grid/adaptive computing

    Increasedend-user

    satisfaction/ExtendedBI reach

    SignificantTCO

    reduction

    HPA S i

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    15/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 15

    HPA Sc enar ios

    Ready for h igh data vo lumes Queries that routinely involve access to many millions of records and

    may involve up to billions of records

    Examples: retail, utilities, telephone companies

    Chal lenging response t im e SLAs

    Example: service level agreements for call center operators demandshort response times for good closure rates

    Unpred ic t ab le t ypes o f quer ies

    Far more different data sets and aggregations than traditionaloptimization and caching strategies can handle

    Excellent response times for any drill-down, slice & dice,

    Examples: on-demand reporting for different user groups,ad hoc analyses

    HPA Impac t on Main tenance

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    16/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 16

    HPA Impac t on Main tenance

    How HPA reduc es ma in t enanc e e f fo r t Aggregate maintenance

    Automated caching strategies

    Building database indexes

    Improves implementation and administration Reduces TCO

    The relational aggregate/caching approach is notobsoletewith HPA, and is suitable for some scenarios:

    Smaller data volumes

    Predictable query patterns

    Relaxed response time constraints

    Caching reduces potential high load on the HPA engine

    HPA Test Resul t s

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    17/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 17

    HPA Test Resul t s

    Pre l im inary Lab Test Resu l t sLab tests with real SAP NetWeaver BI customer data

    MultiProvider with 9 InfoCubes

    About 850 million records together Customer aggregates used

    The 7 most important, critical queries tested

    Improvement factor 25

    Hardware example:

    At SAPPHIRE 2005 in Boston, SAP Executive Board member Shai Agassi demonstrated the

    new capability for high performance analytics by launching complex queries against abillion data records and getting results back in seconds

    For the demonstration, the HPA engine ran under 64-bit Linux on 8 blades, each with dual3.6 GHz Intel Xeon processors and 8 GB of RAM, mounted above a HewlettPackardProLiant 640 server with 4 Intel Itanium processors running SAP NetWeaver 2004s BI under

    HP-UX, and an HP StorageWorks filer, all in a standalone cabinet

    BI Perform anc e Tec hniques : Speed vs Re usabi l i t y

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    18/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 18

    BI Perform anc e Tec hniques : Speed vs. Re-usabi l i t y

    HPA of fers h igh re-usabi l i ty t oget her w i t h fast per formanc e in

    a manner not possib le w i th o t her tec hn iques!

    Canned Report s

    Query Cac he

    Summ ary Tables

    Database

    HPA

    Less

    fast

    Fastest

    Performance

    Lowest Highest

    Re-usability

    High re-usability Very few specific objects (or sets) can be used toaccelerate many different queries

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    19/64

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    20/64

    Overview

    HPA Configuration

    HPA Administration

    HPA in Detail

    Summary

    Appendix: Some Performance Features 2004s

    HPA Inst a l la t ion

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    21/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 21

    HPA Inst a l la t ion

    Ins t a l la t ion for HPA in a box

    During ramp-up of SAP NetWeaver 2004s, the underlying TREXsoftware for HPA will only be available preinstalled in a box

    The box may be standalone or fit into an existing customer rack

    The box will contain blade servers with 64-bit Intel processors

    The OS for the blades is Linux SLES 9

    The blades will be delivered with the complete HPA engine

    preinstalled During the ramp-up phase, the box will contain Intel Xeon CPUs in

    a choice of Hewlett-Packard or IBM hardware

    Sizing for HPA (Ram p-Up Delivery)

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    22/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 22

    Sizing for HPA (Ram p Up Delivery)

    Sm al l T-shir t Size

    20 parallel user sessions

    250 million rows total

    500 byte / row

    Mediu m T-shir t Size

    50 parallel user sessions

    500 million rows total

    500 byte / row

    Large T-shir t Size

    100 parallel user sessions 1,000 million rows total

    500 byte / row

    Intel-based blade technology under 64-bit Linux SLES9

    HPA Index Main t enanc e 1

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    23/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 23

    HPA Index Main t enanc e 1

    Ac c ess ing HPA Index Ma int enanc e

    Similar to aggregatemaintenance viacontext menu for aparticular InfoCube

    Or direct access viatransaction RSDDV

    HPA Index Main t enanc e 2

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    24/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 24

    de a e a c e

    Conf igur ing RFC Dest inat ion for in t he BI syst em

    Configure HPA setting to the SM59 setting(this needs only be done once)

    HPA Rol l-up and Change Run

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    25/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 25

    p g

    Roll-up

    Update of newly loaded data to HPA index

    Similar to roll-up process for aggregates

    But there is only one HPA index to adjust

    So the process is much easier and faster

    Same process for HPA and aggregates: process chains do not need to bechanged after implementing HPA

    Change Run

    Update HPA index after master data/hierarchy changes

    HPA index has same granularity as the InfoCubeNo c hanges needed fo r t ransac t ional da ta

    HPA index includes some master data tables which have to be updated within

    minimal runtime Adjustment of HPA index for key date changes of time-dependent attributes is

    not necessary

    Same process for HPA and aggregates: process chains do not need to bechanged after implementing HPA

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    26/64

    Overview

    HPA Configuration

    HPA Administration

    HPA Engine in Detail

    Summary

    Appendix: Some Performance Features 2004s

    Demo

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    27/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 27

    Demo

    Zero Admin is t ra t ion

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    28/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 28

    HPA Goal : Zero Adm in is t ra t ion

    HPA will feature automatic

    Monitoring

    Configuration

    OptimizationSelf-repair

    TREX Alert Server

    Preconfigured checks

    New checks can be added

    Feeds any alerts to BI

    TREX Admin Tool

    Offers all admin options

    Access to relevant optionsvia BI planned

    Any alerts shownin portal view

    SAP NetWeaver BI

    TREX

    HPAEngine

    HPA Arc h i t ec t ure : H igh Avai lab il i t y and Sc alab i l i ty

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    29/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 29

    RFCServerRFCServer

    IndexServerBackup

    HPAEngine

    Index

    ServerMaster

    IndexServer

    Primary

    HPAEngine

    Index

    ServerCohort

    IndexServerNode

    HPAEngine

    Index

    ServerCohort

    IndexServerNode

    HPAEngine

    Name

    ServerMaster

    NameServer

    Primary

    NameServer

    Backup

    Filer

    Index

    ServerCohort

    HPA Fast Index Update

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    30/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 30

    Queries Index updates

    TREX

    Main Index

    Delta Index

    Benefits: Smaller files

    updated

    Faster replication

    HPA Engine

    Queries runagainst both

    indexes Merge delta indexes

    as background job Triggered by size Per schedule Or manually

    TREX Adm in is t ra t ion Tool

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    31/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 31

    The TREX admin tool can be used for many tasks

    The task shown here is landscape configuration

    Aler t s in TREX Adm in Tool

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    32/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 32

    Click here to view alerts

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    33/64

    Overview

    HPA Configuration

    HPA Administration in TREX

    HPA Engine In Detail

    Summary

    Appendix: Some Performance Features 2004s

    HPA Index Bui ld T ime

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    34/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 34

    SAP NetWeaver BI High PerformanceAnalytics Capability

    Data Acquisition

    InfoCubes

    BI AnalyticEngine

    BusinessExplorer

    Any Tool

    Parallel indexing of InfoCube datavia standard BI processes

    Vertical decomposition& compression to minimize

    memory footprint of HPA indexes

    AnySource

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    35/64

    HPA Query Run T im e

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    36/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 36

    SAP NetWeaver BI High PerformanceAnalytics Capability

    Data Acquisition

    InfoCubes

    BI AnalyticEngine

    BusinessExplorer

    Any Tool

    AnySource

    Aggregation on the fly

    Merging and resultspreparation for BI queries

    HPA Arc h i tec t ure: TREX In dex Server and HPA Engine

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    37/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 37

    HPA Engine Indexes BI InfoCubes

    Materializes joins

    Aggregation on the fly

    Index Server

    HPAEngine

    Indexes

    TREX

    RFC

    TREX Index Server

    Indexes and searches

    Contains HPA engine

    Why HPA is so fastVertical decomposition

    Horizontal partitioning

    Smart compression

    HPA Arch i t ec t ure : HPA Ver t ic a l Dec om posi t ion

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    38/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 38

    HPA

    stores tables by column

    To find all instances of an attributevalue:

    Go to the attribute column

    Read its row values

    Classic a l DB

    stores tables by row

    To find all instances of an attributevalue:

    Go to the first row

    Check the attribute value Go to the next row

    Check the attribute value

    Repeat for each row in the table

    Att1Att1

    Tuple2

    Tuple1Att2

    Att2Att2

    ...

    If you dont havea predefined indexready you can wait

    a long time

    HPA Arc h i tec t ure: HPA Hor izonta l Par t i t ioning

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    39/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 39

    Fact TableFact Table

    PhysicalIndex

    Server

    IndexPart 1

    PhysicalIndex

    Server

    IndexPart 2

    PhysicalIndex

    Server

    IndexPart N

    LogicalIndex

    Server

    IndexMetadata

    Storage

    TREX Services

    Index

    Part 1

    Part 2

    ...

    Part N

    Metadata

    Fact Table

    Part 1

    Part 2

    ...

    Part N

    LargeTables

    HPA Arc h i t ec t ure : Sm ar t Com press ion

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    40/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 40

    Compress ion

    Dictionary-based compression

    Size of table keys significantly reduced

    Lab Resul t s Lab results on real data have shown compression factors up to 20

    The HPA engine can handle huge data volumes in main memory without need for costly disk accesses

    HPA Arc h i tec t ure: HPA Eng ine

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    41/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 41

    HPA Queri es

    Automatically join required indexesas specified in the HPA index

    Aggregate large amounts of data inparallel

    Restrict aggregated rows withboolean query expressions

    Process more than one booleanexpression at once

    Very fas t aggregat ion a lgor i thm

    Exploits integer coding formax speed and min I/O

    Designed to run on latest blade

    servers and grid landscapes

    Optimizes usage of memory andcache resources

    Optimized for BI

    Compressed m em ory index

    Many million documents can fit in1 GB memory

    Stores by column (not by row)

    Can split indexes horizontally for

    parallel processing

    Hybr id dynamic index

    Enables indexing of huge volumesof structured data

    Keeps only required columns inmemory

    HPAEngine

    TREX

    Engine

    Engine

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    42/64

    Overview

    HPA ConfigurationHPA Administration in TREX

    HPA In Detail

    SummaryAppendix: Some Performance Features 2004s

    HPA Sum m ary

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    43/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 43

    Query proc ess ing in mem ory

    Sophisticated compression

    Horizontal and vertical data partitioning

    Enabled for grid and adaptive computing

    Based on 64 bit architecture

    HPA is t ransparent fo r t he app l ic a t ions and 3rd par ty t oo ls

    Available for all applications and tools that

    Use SAP NetWeaver BI for reporting and analytics

    Are based on SAP NetWeaver 2004s

    Easy migration from existing aggregate setup to HPA

    Queries automatically use HPA

    Data maintenance via BI, minimum admin for HPA

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    44/64

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    45/64

    Overview

    HPA ConfigurationHPA Administration

    HPA In Detail

    SummaryAppendix: Performance Features 2004s

    OLAP Cac he Enhancem ent s 1

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    46/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 46

    Delt a Cac hing

    A query reads data from the OLAPCache as far as possible

    Additional requests will be readfrom the InfoCubes (F) fact table

    These additional requests willbe updated into the OLAP Cache

    Restrictions

    If the InfoCube is compressed, delta cannot be determined Cache entrywill be built up completely

    OLAP Cache

    Request 1

    InfoCube (F)

    Request 2

    Request 3

    Request 1

    Request 2

    Request 3

    Request 4

    Request 5

    OLAP Cac he Enhancem ent s 2

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    47/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 47

    Delt a Cac hing

    Setting per query

    Transaction RSRT

    Activation of deltacaching

    OLAP Cac he Enhancem ent s 3

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    48/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 48

    Cac hing for Mul t iProvider Quer ies

    Invalidation of the basis InfoProviders depends on the groupingtype setting

    No InfoProvidergrouping: complete query

    result is kept and will bediscarded if one if theInfoProvider changes

    Grouping: one group storesall InfoCube results, other

    objects in one or moregroups (uses deltacapabilities of infoCubes)

    Grouping with

    separate basisInfoCubes: every InfoCuberesult is stored separately

    Separation of allInfoProviders

    Query Ex ec ut ion

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    49/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 49

    Query Ex ec ut ion

    All query splits can beparallelized

    Splits of MultiProviderinto InfoProviders

    Splits of InfoProviderinto Aggregates

    Splits ofAggregates/InfoProviderinto E and F tables

    Splits in analytical engine (constant selection, cell calculations) areparallelized with the exception of non-cumulative calculations

    Number of parallel processes = min (6, max (0, #free processes 3))

    not more than 6 parallel processes

    F E

    MultiProvider

    InfoProviders

    Aggregates

    F/E tables

    Rol l -up and Change Run 1

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    50/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 50

    Roll -up and Change Run

    Parallelized across aggregates

    Sequence (with regards to aggregate hierarchy) is obtained

    Several change runs can run in parallel if there is no common

    InfoCube Settings can be made in the

    Batch Manager Framework(transaction RSBATCH)

    Rol l -up and Change Run 2

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    51/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 51

    Ex am ple for para l le l c hange run/ro l l -up

    First step: calculate all basis aggregates (if there are more processesavailable than basis aggregates, not all processes are used)

    Once one job is finished, the system checks for the next level ofaggregates

    The following graphic is simplified as not all jobs end at the sametime some jobs might be entangled

    3 available parallel

    processes

    Step 1InfoCube

    Basis

    Aggregates

    Aggregates

    BA1 BA2

    A1 A2 A3 A4

    BA1 BA2

    Step 2A1 A2 A3

    Step 3 A4

    Compression

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    52/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 52

    Compression

    General Optimizations

    Inserts will be used automatically in case InfoCube is empty or therespective time value is new

    Automatic DB parallelism via MERGE/UPSERT (also implemented in

    DB6)

    Generation of compression programs has been reducedperformance improvement especially for small requests

    OLAP Cac he Enhancem ent s 4

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    53/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 53

    Selec t ive Dele t ion

    Deleting in partitions

    System automaticallydetects a databasepartition which matches

    a subset of the recordsto be deleted

    Scenario: you archive abunch of records and

    expect performanceimprovement bydropping completepartitions

    Recommendation: usethis flag for large datasets; for small sets theselective deletion might

    work faster than the initial check for partitions

    Par t i t ion ing

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    54/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 54

    Enhanced part i t ion ing

    Additional database platforms support partitioning

    DB2 / UDB: Multi-dimensional Clustering

    Clusters can be built out of several keys (e.g. month and org unit)

    Benefits: improved query performance, small indexes, fast insertion/deletion, less

    table reorganization Supported tables: PSA, Fact tables and DataStore object tables

    Further SAP optimizations: automatic statistics (no job needs to be scheduled),automatic tablespace enhancements

    DB2 / AS400: Range Partitioning

    Maximum 1000 partitionsMS SQL Server: Range Partitioning

    Fact tables (E/F), PSA

    ORACLE

    Partitioning is not used if partitioning option is not installed this is necessary forSAP NetWeaver installations without usage type BI

    Repart i t i on ing 1

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    55/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 55

    Repar t i t ion ing

    Scenarios: after the initial partitioning you want someadditional partitions; after archiving data you want to mergesome earlier partitions

    Three use cases of repartitioning tool

    Append new partitions

    Merge partitions

    Complete repartitioning

    Repart i t i on ing 2

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    56/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 56

    Repar t i t ion ing

    Append new partitions

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    57/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 57

    Repar t i t ion ing

    Merge partitions

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    58/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 58

    Repar t i t ion ing

    Complete repartitioning

    MonthFiscal Year Variant

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    59/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 59

    Dat a Load Per formanc e Improvem ent s

    New Data Transfer Process

    High scalability by enhanced parallelism

    DataStore Object

    Write-optimized DataStore object

    Improved activation process

    Improved rollback process

    Centralized Monitoring

    via Administration Cockpit Pro-active support of SAP

    NetWeaver BI administrator instatus tracking andperformance optimization(data load, reportingand planning)

    Enhanced data load & queryperformance statistics

    SAP Net Weaver BI Near-L ine St orage St rat egy

    AccessFrequency

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    60/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 60

    q y

    Age of Data

    Online

    Near-Line Storage

    Near-Line St oragein SAP NetWeaver 2004s BI

    SAP NetWeaver BI analyses havedirect access to archived data in

    various storage media

    Availability of historic data whilereducing costs

    Physical decoupling of frequentlyused and less frequently or rarely

    used data

    Reloading of data into theInfoCube or DataStore Objectonly necessary in exceptional

    cases

    Furthe r Informat i on

    P bl i W b

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    61/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 61

    Publ ic Web:www.sap.com

    SAP Developer Network: www.sdn.sap.com Business InformationWarehouse SAP BW Performance Tuning Knowledge Center

    SAP Customer Services Network: www.sap.com/services/

    NetWeaver Developers Guide: www.sdn.sap.com/sdn/developersguide.sdn

    Re la ted Work shops /Lec t u res a t SAP Tec hEd 2005IM251, Enterprise Data Warehousing Enhanced Administration &Monitoring in SAP NetWeaver 2004s, 2h Hands-on

    IM252, Enterprise Data Warehousing Enhanced Modeling Capabilities inSAP NetWeaver 2004s, 2h Hands-on

    Relat ed SAP Educ a t ion Tra ining Oppor tun it ieshttp://www.sap.com/education/

    Quest ions?

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    62/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 62

    Q& A

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    63/64

    SAP AG 2005, SAP TechEd 05 / IM205 / 63

    Please complete your session evaluation.

    Be courteous deposit your trash,and do not take the handouts for the following session.

    Feedback

    Thank You !

  • 8/7/2019 Hi Perf Analytics in SAP NetWeaver 2004s

    64/64