Driving Peak Performance 1
Session 2: Driving Peak Performance Abstract We know you want the fastest performance possible for your deployments, and yet that relies on many choices across data storage, query approaches and hardware usage. To help those who don t know where to start, and those looking to learn new approaches, this session will discuss the decision-points to implementing a performance architecture, use of in-memory and latest query technology, as well as optimized workload systems. 2
Agenda Performance Matters BA: Architected for Performance Beyond BA: The IBM Advantage 3
Evolution of Business Intelligence drives performance 4
Sustained Performance Investments Investment Areas Query and In-Memory Advancements Hardware/Software Optimizations Best Practices Recommendations 5
Sustained Performance Investments Results IBM Cognos BI V8.4 V10.1 IBM Cognos BI V10.1 V10.2 56% faster overall 27% faster overall Overall 67% faster since V8.4 Same applications Same tests 6
Succeeding with Analytics: Supply Meets Demand Performance Matters IBM Cognos Business Intelligence meets Performance 7
IBM Cognos Business Intelligence Reports Dashboards Ad-hoc Query Real-Time Monitoring Microsoft Office Integration Analysis & Exploration Mobile Devices 8
Cognos Business Intelligence Platform APPLICATIONS ABOVE PLATFORM Performance Scalable Secure Open Data Access Extensible Flexible Central Administration DATA TIER BELOW 9
Dynamic Query Layer Dynamic Query & Dynamic Cubes Integrated within Query Layer All Interfaces Dynamic Query Dynamic Cubes Compatible Query Open Data Access Aggregate Awareness In-memory Acceleration 10
Succeeding with Analytics: Supply Meets Demand Dynamic Query Transformation Engine SQL MDX Caches Database Aggregates Data Warehouse Data Warehouse 11 Database
Aggregate Advisor Easy performance improvements Wizard interface for easy optimization In-memory aggregates for turnkey performance improvement Database aggregates to enhance data warehouse 12
Succeeding with Analytics: Supply Meets Demand Demo Aggregate Advisor 13
Succeeding with Analytics: Supply Meets Demand 14
University of Colorado Dynamic Cubes helps us turn Cognos from a packaged reporting engine into a self-service BI engine With Dynamic Cubes, performance will continue to be fast even as our data volumes grow After running the Aggregate Advisor, a report that used to take over 90 minutes ran in 3 seconds Molly Doyle Assistant Director for IRM University Information Systems University of Colorado, Office of the President 15
Dynamic Cubes: Recap High performance analytics on large data volumes Comprehensive in-memory caching Accelerated with aggregate awareness Aggregate advisor that recommends and applies optimized aggregates without re-authoring Fully integrated with the Cognos BI platform 80x Faster on average when leveraging in-memory aggregates 16
Succeeding with Analytics: Supply Meets Demand Beyond BA : The IBM Advantage 17
Power Analytics Solutions Balanced design to handle complex queries and analytics Bring your analytics to your data: 70% of the data used for analytics originates on zenterprise Easily delivers on modern analytics requirements for: Timely, accurate and secure Superior availability, scalability and performance Rapid deployment and expansion Reduced cost and complexity Evolves with your business: Start where you want and grow without re-architecting 18 Data In Business Analytics Business Intelligence Insight Out Predictive Analytics Data Warehousing Data Warehouses Operational Data Stores Data Marts Business System / OLTP When to consider.
Power Analytics Solutions Balanced design to handle complex queries and analytics Extensive family of servers Workload optimized systems and software Fast time-to-value for popular IT solutions with factory pre-integration Synergy with DB2, Websphere and Business Analytics (Cognos and SPSS) When to consider 19
IBM Business Intelligence on PureApplication Traditional Manual install Weeks to usability Multiple management interfaces Static hardware Repetitive maintenance tasks PureApplication Up and running in minutes Automated deployment Built-in workload elasticity Single point of management Virtualized Single entry point for updates Value Faster time to value Automated workload scalability Management & monitoring Lower cost of ownership When to consider 20
Speed of business analytics (PLACEHOLDER HIDE SLIDE) TIL 4/3 Cognos BI with BLU Acceleration Create Tables, Load and Go! Instant, orders of magnitude performance boost for speed of business analysis Handles terabytes of data No indexes or aggregates to create or tune Multi-platform software flexibility 21 Poor Performing Oracle or Teradata Warehouse Easily create & load a BLU in-memory mart Cognos BI with BLU Acceleration Slice & dice OLAP Multi-platform software
Highlights Performance Matters BA: Architected for Performance Beyond BA: The IBM Advantage 22
Things You Can Do: Research of Cognos Dynamic Cubes Upgrade from Cognos BI V8.x to 10.2 Streamlined process Lifecycle Manager Multi-version Coexistence Cognos Business Intelligence: http://www-01.ibm.com/software/analytics/cognos/ Dynamic Cubes RedBook: http://www.redbooks.ibm.com/abstracts/sg248064.html Upgrading Cognos: http://www-01.ibm.com/software/data/cognos/customercenter/upgrade.html zenterprise: http://www-03.ibm.com/systems/z/hardware/zenterprise/index.html Power 7+: http://www-03.ibm.com/systems/power/hardware/index.html PureApplications: http://www.ibm.com/ibm/puresystems/ca/en/ 23
Thank You! 24
OLAP Technology 25 Application Objective Key Question Write-back, what-if analysis, planning/budgeting, or other specialized applications? Can the source be a data warehouse that is structured in a star/snowflake schema? Must the application source be one or several operational/transactional systems, and is a consistent interactive analysis experience a top priority for your users? Must the application source be one or several operational/transactional systems, and is there a need to control latency (ie, some queries hitting the cache / some queries hitting latest data)? If yes TM1 Dynamic Cubes PowerCubes OLAP Over Relational (OOR) Notes / Considerations Medium data volumes High volatility / Write-back Note: aggregation happens on the fly, which can impact performance at high data & high user volumes High data volumes Low latency / Fast performance Optimized aggregates / aggregate-aware Note: Star or snowflake schema is the optimal structure for reporting Highly recommended to maximize performance Low / medium data volumes Data movement into cube structure Note: Data latency is inherent to cube build times Data volume per cube must be managed Low / medium data volumes Caching for performance (Dynamic Query) Leverages existing Framework Manager model Note: Processing associated with operational/transactional systems impacts performance
26