Configure Oracle Hybrid Columnar Compression to Optimize Query Database Performance up to 10x

Configuring HCC for OptimizingDatabase Query Performanceby up to 10x Insert Picture HereCON6528Art LichtChief Systems Architect, Oracle Solution CentersEnterprise Solutions Group 2Copyright © 2012, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 13

Program Agenda Compression Overview Configuration Best Practices Case Study Real World Experiences Getting Started with Compression3Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Data Growth Challenges •Exponentially growing amounts of data – Explosion in online access and content – Government data retention regulations • Performance often declines as data balloons • IT budgets are flat or decreasing • Need to grow data – Without hurting performance – Without growing cost – Powerful and efficient compression is key4Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Storage Array Compression• Works with any application• Application is unaware that the data is stored compressed• Data must be uncompressed to be transferred to the application server• Compression rates vary between 2-10x depending on data and algorithm• Can improve write performance• Can slowdown read performance6Copyright © 2012, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 13

Database Aware Compression Types Table Direct-Path Insert Notes Create/Alter Compression Table Syntax Method Basic Compression COMPRESS Rows are Does not maintain [BASIC] compressed with compression for Data basic compression Manipulation Language (DML) INSERT/UPDATE operations after bulk load OLTP Compression COMPRESS Rows are Maintains compression FOR OLTP compressed with for DML operations. OLTP compression Hybrid Columnar COMPRESS Warehouse Ideal for active data Compression FOR QUERY compression warehouses or for use (Warehouse Query) [LOW|HIGH] in other environments with partitioning Hybrid Columnar COMPRESS Archive Ideal for archive/historic Compression FOR ARCHIVE Compression data (Archive) [LOW|HIGH]7Copyright © 2012, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 13

Database Aware Compression• Data is compressed by the database• Data remains compressed as it’s transferred between storage and compute• Transparent to the application• Compression rates vary between 2-50x depending on data and algorithm• Can significantly improve query performance• OLTP Table Compression is a feature of the Advanced Compression Option (ACO), a separately licensed database feature• Hybrid Columnar Compression (HCC) is included in Oracle Database 11g Release 2 Version 11.2.0.3 and available exculsivy with Oracle storage (Exadata, Sun ZFS Storage Appliance and Pillar Axiom)8Copyright © 2012, Oracle and/or its affiliates. All rights reserved. Insert Information Protection Policy Classification from Slide 13

About Hybrid Columnar Compression Compression Unit • Hybrid Columnar Compressed Tables – New approach to compressed table storage – Compressed tables can still be modified using conventional DML operations, such as INSERT and UPDATE – Useful for data that is bulk loaded and queried • Update activity is light • How it Works – Tables are organized into Compression Units (CUs) • CUs are larger than database blocks10x to 15x – Within Compression Unit, data is organized by columnReduction instead of by row • Column organization brings similar values close together, enhancing compression 9Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Warehouse and Archive Compression Query Low/High Archive Low/High •10x average storage savings •15x average storage savings 10x reduction in Scan IO Up to 70x on some data •For cold or historical data Optimized for Speed Optimized for Space Smaller Database Reclaim 93% of Disks Faster Performance Keep Data Online10Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Warehouse Compression Levels •Two levels of compression: LOW and HIGH –HIGH typically provides a 10-15x reduction in storage –LOW typically provides a 5-10x reduction •Reduction in IOs is directly related to the data compression rates achieved –LOW should be chosen for environments where load time service levels are more critical than query performance11Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Database Architecture Database Partitioning and HCC ORDERS TABLE (7 years) 2006-2010 2011 2012 75•50 percent Less percent Historical Data 20 percent Active Data 5 percent Hot Data ActiveSun ZFS Storage Appliance / Existing Storage Pillar Axiom Storage•1 •Copyright © 2012, Oracle and/or its affiliates. All rights reserved. 3

Reduce Storage Costs 50TB -> 15TB ORDERS TABLE (7 years) 2005-2010 2011 2012 75% Archive Data 20% Active Data 5% Hot Data 1.2TB 10TB •50 percent Less Active 2.5TB Sun ZFS Storage Appliance / Pillar Axiom Existing Storage •Copyright © Oracle and/or its affiliates. All rights rights14Copyright © 2012, 2012, Oracle and/or its affiliates. All reserved. Insert Information Protection Policy Classification from Slide 13 reserved.

Customer Findings Compress for Compress for Compress for 2GB OLTP Query High Archive High 4.7 20 23.4 Charge Table Compress for Compress for Compress for 50GB OLTP Query High Query High Payment 2.7 8.1 11.1 Table18Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

HCC Real World Results Retail Trials done using the largest database table from 10Financial very large companies* Average compression ratio was 13xTelecom* Companies with average revenue >$60B USD 19Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Space Reduction Tablename Uncompressed HCC Blocks Space Blocks (8k) Used (8k) Savings PartSupp 20851 4021 5.1X Part 19146 3522 5.4X Supplier 20881 6534 3.1X Orders 16338 3610 4.5X • Overall space savings of 4.21x • Database went from 118GB to 28GB•2 •Copyright © 2012, Oracle and/or its affiliates. All rights reserved. 0

Query Performance Query Without HCC Query Faster HCC High (mins:secs) (mins:secs) Q1 31:21 9:58 3.1X or 68% Q2 8:32 2:05 4.1X or 78% Q3 14:01 3:35 4.0X or 74% Q4 30:52 8:11 3.8X or 72% Scale 100•2 •Copyright © 2012, Oracle and/or its affiliates. All rights reserved. 1

Compression AdvisorOracle Compression Advisor provides an estimate of the compression ratio thatcan be realized• For databases prior to Oracle Database 11g Release 2 use the DBMS_COMP_ADVISOR package• For Oracle Database 11g Release 2 will use the DBMS_COMPRESSION package that is included with the database• There are no specific hardware requirements• Directions and software are located here http://www.oracle.com/technetwork/database/options/compression/compression-advisor-095705.html23Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Compression Advisor Sample DBMS_COMPRESSION.get_compression_ratio ( scratchtbsname => USERS, ownname => TEAM01, tabname => DWB_RTL_TRX, partname => NULL, comptype => DBMS_COMPRESSION.comp_for_query_high, blkcnt_cmp => l_blkcnt_cmp, blkcnt_uncmp => l_blkcnt_uncmp, row_cmp => l_row_cmp, row_uncmp => l_row_uncmp, cmp_ratio => l_cmp_ratio, comptype_str => l_comptype_str ); DBMS_OUTPUT.put_line (l_blkcnt_cmp= || l_blkcnt_cmp); DBMS_OUTPUT.put_line (l_blkcnt_uncmp= || l_blkcnt_uncmp); DBMS_OUTPUT.put_line (l_row_cmp= || l_row_cmp); DBMS_OUTPUT.put_line (l_row_uncmp= || l_row_uncmp); DBMS_OUTPUT.put_line (l_cmp_ratio= || l_cmp_ratio); DBMS_OUTPUT.put_line (l_comptype_str= || l_comptype_str);24Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Enabling HCC with Pillar Axiom1) Upgrade to Oracle Database 11g Release 2 Version 11.2.0.3Refer to My Oracle Support Note 10404530 at support.oracle.com At support.oracle.com, search for and then download and apply patch 13041324.2) Start the database.3) Set the compatibility attributes of the database to 11.2.0.3.0: $ sqlplus / as sysasm SQL> alter system set compatible=11.2.0.3.0 scope=spfile sid=*; SQL> alter diskgroup data_hcc set attribute compatible.asm=11.2.0.3.0; SQL> alter diskgroup data_hcc set attribute compatible.rdbms=11.2.0.3.0;4) Set the storage.type attribute to a value of AXIOM: SQL> alter diskgroup data_hcc set attribute storage.type=AXIOM;Note: Here is a way to check that the disk group is really Pillar Axiom storage. If you execute the previouscommand on non-Pillar Axiom storage, error messages similar to the following are displayed:ORA-15287: could not set disk group attribute storage.type due to incompatible disksORA-15285: disk /dev/mapper/XXXXXXXX violates disk group attribute storage.type 25Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Enabling HCC on the Sun ZFS Storage Appliance1) Upgrade to Oracle Database 11g Release 2 Version 11.2.0.3 Refer to My Oracle Support Note 10404530 at support.oracle.com At support.oracle.com, search for and then download and apply patch 13041324.2) Mount one or more ZFS shares on your Oracle Database server.The following is sample fstab entry for an NFS-mounted ZFS share.Note that when Direct NFS (dNFS) is used, the database sets up the rsize/wsize parameters, as needed.<Appliance IP address>:<Appliance share mount point> nfs nfsvers=3,proto=tcp,hard,intr,rsize=<#bytes>,wsize=<#bytes>3)As user oracle, enable dNFS on the Oracle Database server, by doing the following: cd $ORACLE_HOME/lib directory4) Remove the symbolic link libodm11.so: rm -f libodm11.so5) Create a new symbolic link to point to the dNFS shared library: ln -s libnfsodm11.so libodm11.so6) Start the database. 26Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Oracle Direct NFS with Hybrid Columnar Compression (HCC) Compress at the app level Leverage fast CPUs Optimized for Oracle Save network bandwidth Increase storage efficiency27Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

The HCC Multiplier Effect10x less storage10x better disk bandwidth10x more data in Flash Cache10x more data in database memory10x smaller test, development databases10x smaller and faster backup and recovery 28Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

For More Information• Attend Sessions: • CON8433 on HCC • CON8439 on Transforming Oracle Database Performance• Read • http://www.oracle.com/technetwork/articles/servers-storage-admin/perf-hybrid-columnar-compression- 1689701.html• View • www.youtube.com/watch?v=rXMmh0foYmE • www.youtube.com/watch?v=2EctIMoAT-Y • www.youtube.com/watch?v=rbZFXOH_z3Y• Learn • education.oracle.com/storage• Visit • Oracle Hardware on Facebook: www.facebook.com/oraclehardware29Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Join the Conversation Like us on Facebook • Engage with Oracle Experts • Share knowledge and experiences • Keep up with hardware topics & trends • Access Oracle Hardware content easily, including events, webcasts, white papers and more Join the Oracle Hardware Hub today facebook.com/oraclehardware30Copyright © 2012, Oracle and/or its affiliates. All rights reserved.

Configure Oracle Hybrid Columnar Compression to Optimize Query Database Performance up to 10x

by Oracle Hardware on Oct 25, 2012

Statistics

Views

Actions

0 Embeds 0

Accessibility

Categories

Upload Details

Usage Rights

Report content

Configure Oracle Hybrid Columnar Compression to Optimize Query Database Performance up to 10x Presentation Transcript