IBM DB2 and SAP

2008 IBM Corporation

DB2@SAP 9 Deep Compression
DB2 LUW Administration for SAP
JongHwan Kim
[email protected]
FTSS, IM Team
SWG, IBM Korea
IBM DB2 and SAP
2008 IBM Corporation 2
Disclaimer & Trademarks
Copyright 2007 IBM Corporation. All Rights Reserved.
Neither this documentation nor any part of it may be copied or reproduced in any form or by any means or translated
into another language, without the prior consent of all of the above mentioned copyright owners.
IBM makes no warranties or representations with respect to the content hereof and specifically disclaims any implied
warranties of merchantability or fitness for any particular purpose. IBM assumes no responsibility for any errors that may
appear in this document. The information contained in this document is subject to change without any notice. IBM reserves the
right to make any such changes without obligation to notify any person of such revision or changes. IBM makes no
commitment to keep the information contained herein up to date.
The information in this document concerning non-IBM products was obtained from the supplier(s) of those products. IBM has
not tested such products and cannot confirm the accuracy of the performance, compatibility or any other claims related to non-
IBM products. Questions about the capabilities of non-IBM products should be addressed to the supplier(s) of those].
References in this publication to IBM products or services do not imply that IBM intends to make them available in all countries
in which IBM operates.
Trademarks
DB2, DB2 Universal Database, AIX, the ebusiness logo, IBM, the IBM logo, xSeries, and the eServer logo are trademarks or
registered trademarks of International Business Machines Corporation in the United States, other countries or both.
A full list of U. S. trademarks owned by IBM may be found at http:// iplswww.nas.ibm.com/wpts/trademarks/trademar.htm.
Linux is a registered trademark of Linus Torvalds.
SAP, R/3 and all other SAP product and service names mentioned herein are trademarks or registered trademarks of SAP AG
in Germany and several other countries.
Oracle is a registered trademark of Oracle Corporation in the United Status and/or other countries.
UNIX is a registered trademark of The Open Group in the United States and/or other countries.
Java and all Java- based trademarks and logos are trademarks of Sun Microsystems, Inc. in the United States and/ or other
countries.
Other company, product or service names may be trademarks or service marks of others.
IBM DB2 and SAP
2008 IBM Corporation 3
DB2 9 Deep Compression
Unit Objectives
After completing this unit, you will be able to:
Understand DB2 deep compression technology and
benefit
Understand how DB2 deep compression feature can be
used in SAP system
IBM DB2 and SAP
2008 IBM Corporation 4
1.1 SAP Databases are growing
Continuously growing (Number of tables, Size)
SAP R/3 4.6D: ~10,000 - 15,000 tables
mySAP ERP 2005: > 50,000 tables
SAP NetWeaver BI
10+ TB databases
TB-size tables
Customers planning for 50, 100, TB
Large SAP customers have hundreds of SAP systems installed
IBM DB2 and SAP
2008 IBM Corporation 5
1.2 DB2 Storage Optimization Solution
Prior to DB2 9
V8 GA - NULL and Default Value Compression
No disk storage consumed for NULL column values, zero length
data in variable length columns and system default values
V8 GA - Multidimensional Clustering (MDC)
Significant index compression can be achieved through block
indexes
One key per thousands of records (vs one key per record with
traditional indexes)
V8 FP4 - Database Backup Compression
Smaller backup images; compress index and lf/lob tablespaces
DB2 9 - Deep Compression
IBM DB2 and SAP
2008 IBM Corporation 6
1.3 DB2 9 Deep Compression example Table GLPCA
Table GLPCA: EC-PCA - 102 columns
RCLNT GL_SIRID RLDNR RRCTY RVERS RYEAR RTCUR RUNIT DRCRK POPER DOCCT DOCNR DOCLN RBUKRS RPRCTR RHOART
800 4751 8A 0 0 1995 DEM S 1 A 21 1 1000 1402 2
800 4752 8A 0 0 1995 DEM S 1 A 21 2 1000 1402 2
800 4753 8A 0 0 1995 DEM S 1 A 21 3 1000 1402 2
800 4754 8A 0 0 1995 DEM S 1 A 22 1 1000 1402 2
800 4755 8A 0 0 1995 DEM S 1 A 22 2 1000 1402 2
BWART BLART

1000,1402 F67t
8A,0,0,1995,DEM, ,S,1,A x01C
800 4751 x01C 21 1 F67t 0
800 4752 x01C 21 2 F67t 0
800 4753 x01C 21 3 F67t 0
800 4754 x01C 22 1 F67t 0
800 4755 x01C 22 2 F67t 0
Compression dictionary
Uncompressed table
Compressed table
Common sequences
of consecutive bytes
in row replaced with
symbol
IBM DB2 and SAP
2008 IBM Corporation 7
1.4 DB2 Deep Compression Concepts
Dictionary based - symbol table for compressing/decompressing data
records
Lempel-Ziv (LZ) based algorithm (static dictionary) utilizing 12bit symbols
Dictionary per table stored within the permanent table object (~75KB in size;
disk+memory)
Data resides compressed on pages (both on-disk and in bufferpool)
Significant I/O bandwidth savings
Significant memory (bufferpool) savings
CPU costs
Rows must be decompressed before being processed for evaluation
IBM DB2 and SAP
2008 IBM Corporation 8
1.5 Compression Dictionaries
Stored as part of table data object
Not accessible externally
Size max. 150 KB (typical size ~75 KB)
Compression Symbols are 12 bits in size
Each dictionary can have up to 4096 compression symbols
Column values and multi column values can be replaced by a single
symbol
DPF: created and stored on each database partition
Data partitioning: created and stored on each data partition
Compression dictionaries are created by offline table REORG or by
the DB2 INSPECT command
IBM DB2 and SAP
2008 IBM Corporation 9
1.6 DB2 Deep Compression Benefits and Considerations
Benefits
Saves disk space (on tablespaces, log files and backup images)
reduced TCO
Minimizes I/O improved performance on I/O bound system
Increases bufferpool hit ratio improved performance
Considerations
Increased CPU workload for compression / decompression,
especially for operations that access the table itself (e.g. table scans,
non-sampled RUNSTATS, REORG)
Database heap - Memory for compression dictionaries is allocated
from database heap (on average ~75 KB per dictionary)
Utility heap - Building compression dictionaries requires a temporary
in-memory buffer of 10 MB that is allocated from the utilities heap
IBM DB2 and SAP
2008 IBM Corporation 10
1.7 Limitations of Data Row Compression
Row compression is only supported for tables supported by table REORG.
The following tables cannot be compressed:
RCT tables (range-clustered tables)
Catalog tables
Declared global temporary tables
System temporary tables
Row compression is only applicable to rows in data objects table space!!
(the key to this is if it is stored with tables, or somewhere else)
Pure XML in separate object XML data is not compressed (can be compressed in V9.5)
LOBs are stored in separate object LOB data is not compressed
LONGs are stored in separate object - LONG data is not compressed
Indexes are not compressed
Compression is not practical for a table smaller than 100KB, since the
space savings may not offset the storage requirements for the dictionary.
Row compression support and table data replication support will not be
compatible.
DATA CAPTURE CHANGES option is not compatible with the COMPRESS YES option. If the
compress row attribute is NO for the table but the table has a dictionary, the DATA CAPTURE
CHANGES option is also not allowed since there may be rows that are compressed in the table.
IBM DB2 and SAP
2008 IBM Corporation 11
1.8 How Do I Compress a Table?
In V9.1, it takes two steps to compress data in a table:
1) The table COMPRESS attribute must be set to YES
CREATE TABLE <table name> COMPRESS YES
ALTER TABLE <table name> COMPRESS YES
2) A Compression Dictionary must exist for the table object
REORG--<table name> KEEPDICTIONARY|RESETDICTIONARY
INSPECT ROWCOMPESTIMATE-TABLE--+-NAME--<table-name>
Once both step 1) and step 2) are satisfied, all data subsequently populated into the
table is subject to being compressed.
IBM DB2 and SAP
2008 IBM Corporation 12
1.8.1 Enabling Compression with V9.1
3)
OFFLINE
REORG
Dictionary
Dictionary
1) CREATE

COMPRESS YES
2) LOAD
Or, ALTER <existing
table>
COMPRESS YES
Or, IMPORT,
INSERT,
IBM DB2 and SAP
2008 IBM Corporation 13
1.8.2 Automatic (Compression) Dictionary Creation (ADC) V9.5
Row Compression supports now Automatic Compression Dictionary
Creation (ADC)
Table compress attribute must be set
Synchronous dictionary creation when table data reaches a certain threshold (appr. 1 MB
system default)
Once the dictionary is created new data will be inserted in compressed format
Works regardless of growth mechanism
INSERT
IMPORT
LOAD
REDISTRIBUTE
No need to worry about creating compression dictionary occurs automatically when data
threshold is reached
No lengthy offline reorg operation required to enable compression
Tradeoffs
Compression ratio depends on the first one MB of data
Depending how representative the first MB of data is, the compression ratio less than
optimal
Slight performance hiccup incurred when threshold is crossed. The transaction which
needs to create the dictionary must pay the cost of creating the dictionary.
IBM DB2 and SAP
2008 IBM Corporation 14
1.8.3 Automatic (Compression) Dictionary Creation (ADC) V9.5
1) CREATE

COMPRESS YES
Once threshold is
reached, dictionary
built automatically,
and subsequent rows
compressed
2) LOAD
Or, IMPORT,
INSERT,
Internal threshold
designed as best
compromise b/w
compression ratio, and
dictionary build speed.
Default ~1 MB
Dictionary
Dictionary
IBM DB2 and SAP
2008 IBM Corporation 15
1.9 SAP Support for Deep Compression Overview
Fully Integrated into DBACockpit (NW 7.0 SP12)
Separate Compression Tool for
Monitoring table sizes
Estimating compression disk space savings
Enabling deep compression
Scheduling INSPECT or offline REORG
SAP Note 980067 Using DB2 9 Row Compression
Description and set of SQL scripts and tools for finding deep compression
candidate tables and enabling compression
SAP Note 905614 R3load support for data row compression
IBM DB2 and SAP
2008 IBM Corporation 16
1.10 NetWeaver 2007 Support Compression Check
IBM DB2 and SAP
2008 IBM Corporation 17
1.11 NetWeaver 2007 Support - Compression Enabling
IBM DB2 and SAP
2008 IBM Corporation 18
1.12 Compression Tool in SAP Note 980067 (1)
For use with 4.6C NW2004s
Tool attached to SAP Note rowcompression.zip
Install tool as an SAP transport
Call compression tool program /ISIS/ZCOMP from transaction SE38
Empty list
IBM DB2 and SAP
2008 IBM Corporation 19
The first step is to get a list of tables that are candidates for compression. The
candidates are selected using the following criteria:
20 largest tables (excluding LOBs) in the system which are:
Not compressed
Not volatile
Do not belong to a SAP cluster or pool table, or DB2 index only table
The row size is larger than the minimum row size for the tablespace page size
Schedule the job
1.13 Compression Tool in SAP Note 980067 (2)
IBM DB2 and SAP
2008 IBM Corporation 20
Candidates identified and listed under the Candidates tab
Select table first
1.14 Compression Tool in SAP Note 980067 (3)
IBM DB2 and SAP
2008 IBM Corporation 21
After tables have been compressed, they are listed under the Compressed tab
1.15 Compression Tool in SAP Note 980067 (4)
IBM DB2 and SAP
2008 IBM Corporation 22
1.16 Deep Compression Support in SAP BW
Special support for SAP BW
SAP Note 906765 - DB2 9 data row compression for SAP BW 3.x
SAP BW 3.0 SP 33, SAP BW 3.1 SP 27, SAP BW 3.5 (SAP NetWeaver 04) SP 19
SAP Note 926919 DB2 9 data row compression for SAP NetWeaver BI 2004s
SAP NetWeaver BI 2004s Support Package (SP) 9
Configuration parameter DB6_ROW_COMPRESSION
When set to YES, PSA, DataStore and Fact tables are created with deep
compression enabled. Use report SAP_RSADMIN_MAINTAIN to set the
parameter
After one or more InfoPackages have been loaded into the tables an
offline REORG or INSPECT is required to create the compression
dictionary
Transaction RSRV offers 2 new tests for checking whether a
compression dictionary exists for SAP NetWeaver BI tables that have
deep compression enabled
Check Deep Compression for InfoProvider tables in DB2
Check Deep Compression for PSA tables in DB2
IBM DB2 and SAP
2008 IBM Corporation 23
1.17 BW Tables Deep Compression Check in Transaction
RSRV
Call transaction
RSRV
Select test
Check deep
compression of
database tables
of InfoProviders
in DB2/UDB or
Check deep
compression of
PSA database
tables in
DB2/UDB
IBM DB2 and SAP
2008 IBM Corporation 24
1.18 To Enable Deep Compression for BW Tables
Select offline
REORG or
INSPECT for
tables
chosen for
compression
IBM DB2 and SAP
2008 IBM Corporation 25
1.19 Customer 1 Experience
Compression test on a real customer BW database
a) Total db size (over all partitions): 3.356 TB
b) Total used space (over all partitions): 2.873 TB
c) Total free space (over all partitions): 480 GB
d) Data Tablespaces with tables, which will be compressed:
Total size: 2.398 TB used: 2.029 TB Free: 0.367 TB
e) Index Tablespaces, which will not be compressed:
Total size: 0.781 TB used: 0.689 TB Free: 0.091 TB
g) Data and Index Tablespaces, which will be not compressed:
Total size: 0.177 TB used: 0.155 TB Free: 0.022 TB
Database size reduced from 2.873TB to 1.453TB (50%)
169 72.8% 6.32 23.24 39573969 10 /BIC/B0000285000
172 66.9% 8.54 25.78 23370458 10 /BIC/B0000326000
43 74.2% 3.05 11.80 19474261 20 /BIC/AZOSDBL0100
105 74.4% 36.18 141.46 191362309 100 /BIC/AZOSDOR0200
162 75.7% 5.23 21.48 77390826 10 /BIC/FZBSDOR02
58 81.5% 13.23 71.49 159259747 100 /BI0/F0PUR_C01
Total Estimated
Disk
Savings
(GB)
Compressio
n ratio
(%)
Sample
Compressed
size (GB)
Sample size
(GB)
Number of
rows
% of
original
table
Table name
IBM DB2 and SAP
2008 IBM Corporation 26
1.20 Customer 2 Experience INTER Versicherung
(Germany)
SAP R/3 4.7 with IS Insurance.
DB size:
Overall DB size (used) uncompressed: 266 GB
Overall DB size (used) compressed: 151 GB (43 % savings)
Compressed size includes larger index sizes due to large RID conversion (for all the tables that were
compressed and therefore reorganized)
SAP note #980067 used to identify tables for compression - 310 tables, shrinking from 132,5 GB to
39,7 GB (70 %)
Perfomance behaviour:
During two weeks of production
Dialog response time for their 30 most important transactions went down from 285 ms average to
198 ms average (30% improvement).
Dialog response time for all transactions went down from 247 ms average to 183 ms average
(26% improvement).
There was also an improvement for batch jobs.
They pay a slight CPU penalty. CPU user time went from 18,3 % average to 19,6 %
average (+ 5%) ... they for sure never had a CPU bottleneck
No performance problems before, overall they ran a well balanced system
IBM DB2 and SAP
2008 IBM Corporation 27
1.21 Customer 2 Experience INTER Versicherung
(Germany)
Week 46 database uncompressed
Week 47 database compressed
Through monitoring the SAP system performance for these two weeks, we can
conclude:
The overall response time had improved by over 23%, for both Dialog and Batch transaction.
The main contributor was the reduced database response time.
For Dialog transactions, bigger throughput was achieved, and with less CPU
For Batch transactions, CPU usage was up by 22.3%
IBM DB2 and SAP
2008 IBM Corporation 28
1.22 Customer 3 Experience - SAP BI PoC Unnamed
customer
PoC done on a customers 20TB database and workload
p595 and DS8300 storage
Total DB reduced from 16TB to 8TB (used pages)
Average Compression Rate
ODS 79% (26 tables)
PSA 64% (517 tables)
Fact 86% (100 tables)
Average Query Times
Transactions/sec 1.39 to 1.44 (3.5%)
Response (sec) 10.6 to 7.12 (23%)
IBM DB2 and SAP
2008 IBM Corporation 29
1.23 Backup Compression and Data Row Compression
Test backup compression in addition to tables with row compression
Backup compression can be expensive and may not provide much added value in
additional savings to backup image size
Time/size/value depends on the percentage of table space content with row compression.
E.g. Are all tables compressed? Are indexes or long data stored in the same table space?
662
198
1028
468
Total User Time
(seconds)
610816
610816
1510400
1510400
Pages Used
4.68
4.68
11.57
11.57
Table Space
Size (GB)
4.2
4.7
4.2
12
Backup Image
Size (GB)
Data Row and
Backup
Compression
Data Row
Compression Only
Backup
Compression Only
No compression
Scenario
IBM DB2 and SAP
2008 IBM Corporation 30
THANK YOU
FOR YOUR ATTENTI ON !
QUESTI ONS SUGGESTI ONS DI SCUSSI ON