Home >Documents >3.ibm cognos-tm1-9.5.2-why-upgrade

3.ibm cognos-tm1-9.5.2-why-upgrade

Date post:03-Jul-2015
Category:
View:264 times
Download:0 times
Share this document with a friend
Transcript:
  • 1. 2011 IBM CorporationInformation ManagementTM1 9.5.2 Why Upgrade?Brian SimpsonProduct Manager Cognos TM1

2. 2011 IBM CorporationInformation Management2Important Disclaimer THE INFORMATION CONTAINED IN THIS PRESENTATION IS PROVIDED FOR INFORMATIONALPURPOSES ONLY. WHILE EFFORTS WERE MADE TO VERIFY THE COMPLETENESS AND ACCURACY OF THEINFORMATION CONTAINED IN THIS PRESENTATION, IT IS PROVIDED AS IS, WITHOUT WARRANTY OFANY KIND, EXPRESS OR IMPLIED. IN ADDITION, THIS INFORMATION IS BASED ON IBMS CURRENT PRODUCT PLANS AND STRATEGY,WHICH ARE SUBJECT TO CHANGE BY IBM WITHOUT NOTICE. IBM SHALL NOT BE RESPONSIBLE FOR ANY DAMAGES ARISING OUT OF THE USE OF, OR OTHERWISERELATED TO, THIS PRESENTATION OR ANY OTHER DOCUMENTATION. NOTHING CONTAINED IN THIS PRESENTATION IS INTENDED TO, OR SHALL HAVE THE EFFECT OF: CREATING ANY WARRANTY OR REPRESENTATION FROM IBM (OR ITS AFFILIATES OR ITS OR THEIR SUPPLIERSAND/OR LICENSORS); OR ALTERING THE TERMS AND CONDITIONS OF THE APPLICABLE LICENSE AGREEMENT GOVERNING THE USE OF IBMSOFTWARE. 3. 2011 IBM CorporationInformation Management3TM1 9.5.2 EnhancementsRead / Write Scalability & Performance Parallel Interaction Data ReservationBI Integration BI Report Performance TM1 iWidgetsContributor Usability Multi-Node Edit Text Wrap Spreading ShortcutsConformance Red Hat Linux* Excel 2010 BI Pkg. Con. / C10Language New LanguagesRules & TI Min, Max, Avg Count, DCountTM1Top Sandbox Job Queue 4. 2011 IBM CorporationInformation Management4Parallel Interaction During periods of read/write contention,writer performance is improved Performance is steady as users scale Analysis of updates in real-time Concurrent data spreading Eliminates Writer Partitioning Best Practice Intra-Day Data Imports do not inhibitwriters Intra-Day TI Processes do not inhibitwriters Faster parallel data loads to same cube Scale leverages server core capacity Job Queue multi-threads processeingGoal Greater scale for read/writeenvironments Faster performance for writers Simultaneous Activities (datamaintenance writes / TI & reads) Maintain reader performance Greater CPU utilizationBenefitsNotes: PI provides benefits when contention exists PI is not targeted at reader performance Meta Data maintenance still blocks Memory consumption will rise ~ 10 30% Watch for core saturation as you scale 5. 2011 IBM CorporationInformation Management5Demonstrating 9.5.1 Locking Behavior a brief reviewCubes linked by rules will lock together - A read toCube A will lock a write to Cube A & BRules create adependencybetween Cube A& Cube BCube ACube BBlocked writers will stack up behind concurrentreaders and wait for the reads to flush out.Readers are free to process together, so multiplereads can occur in parallel to all cubesWhen the cube is free, the writers process serially(they block each other)ReadersWriters 6. 2011 IBM CorporationInformation Management6Without Concurrency, 9.5.1 performance will be on par with 9.5.2 PIThis is a TM1 9.5.1 example demonstrating behaviorwithout concurrency conflictsThese readers and writers are not conflicting with eachother, therefore there is no waiting occurring due toobject locking9.5.2 with PI will not improve performance of thisscenario because there is no concurrency conflict 7. 2011 IBM CorporationInformation Management79.5.2 PI Enhances Writer Performance during High Concurrency No Waiting!TM1 9.5.2 PI removes object locking that occurs asa result of data reads or writesWriters are no longer blocked by readers (orother writers) they process without delay*9.5.2 PI performance improvement vs. 9.5.1 is noticedwhen in 9.5.1 scenarios demonstrating contention - andwhen there are sufficient Server Cores to handle thegreater level of concurrent transactions* 9.5.2 still has object locking scenarioscaused by Meta Data updates (includingDynamic Subsets, element updates, etc..).. Well fix that in another release 8. 2011 IBM CorporationInformation Management8Existing TM1 Read ConcurrencyCube ARead 1 Cube ARead 2 Cube ARead 3 Cube AProcessing TimeCompletion 9. 2011 IBM CorporationInformation Management9Existing TM1 Read/Write ConcurrencyCube ARead 1 Cube ARead 2 Cube ARead 3 Cube AWrite 1 Cube AWrite 2 Cube AProcessing TimeCompletionWait Time 10. 2011 IBM CorporationInformation Management10Read/Write Concurrency with Parallel InteractionCube ARead 1 Cube ARead 2 Cube ARead 3 Cube AWrite 1 Cube AWrite 2 Cube A Read 2 will not include data impact from Write 1 (because itbegins before Write 1 completes) Read 3 will include data impact from Write 1 (because itbegins after Write 1 completes), but not Write 2 11. 2011 IBM CorporationInformation Management11How does PI work?Cube ARead 1 Cube ARead 2 Cube AWrite 1 Cube ARead 3 Cube AWrite 2 Cube Att+1t+2Time line TM1 previously manage a singleData Tree to Access Cube Data Reads could share a Tree, but Writeshad to wait for Exclusive Access toupdate the Tree Parallel Interaction creates a newData Tree Access Point version foreach Write, allowing concurrentReads & Writes Reads access the latest Data TreeAccess Point to get the most recentupdatesAccess trees to cubedata are versioned,then updated 12. 2011 IBM CorporationInformation Management12Starwood Read/Write Concurrency Test25.4 37.3189.7136.7499.82003.40500100015002000250 Users 500 Users 1000 UsersTM1 ReleaseAVGAggregateResponseTime9.5.2 9.5.1Read / Write Concurrency Improvements 9.5.1 -> 9.5.2 PIWindows 2003 / 8 CoreLoad Runner / Cube Views2 Hour Test9.5.1 -> 9.5.2 Optimizations C-Lock, Data Spreading, and AutoRecalc (very applicable to this test)Servers 8 coresmaxed out @1000 usersPI runs93% FasterPI runs81% FasterPI runs90% Faster 13. 2011 IBM CorporationInformation Management13010203040506070809010000:0000:0400:0800:1200:1700:2100:2500:2900:3400:3800:4200:4600:5100:5500:5901:0401:0801:1201:1601:2101:2501:2901:3301:3801:4201:4601:5001:5501:5902:03Duration of test (hh:mm)%ofTotalCPU0102030405060708090100:00:17:34:51:08:25:42:59:16:33:50:07:24:41:58:16:33:50:07%ofTotalCPU01020304050607080901000:000:080:170:250:340:420:510:591:081:161:251:331:421:501:592:082:162:252:332:422:502:59Duration of test (hh:mm)%ofTotalCPU1000 User R/W500 User R/W250 User R/W9.5.29.5.19.5.29.5.19.5.29.5.1Lower Volumenot taxing 8core server500 User testreaches CPUcapacity1000 Userssaturates server -delays due toCPU constraints 14. 2011 IBM CorporationInformation Management14Parallel Interaction Notes No benefit / possible negative impact to Read performance Reads now compete with Writes for system resources Availability to cache is lessened (due to more frequent write invalidation) Separate Reader environments still a Best Practice Read Only environments operate with PI disabled CPU utilization will be greater, raising importance of server core capacity PI performance benefits are reduced when server CPU power is fully utilized Increase cores in alignment with concurrent read/write scale Dimension Updates still wait for Reads and block other Dimension Updates Isolate Meta Data Updates from Data Updates in TI Processes Other activities remain subject to blocking Save Data All Views with Dynamic Subsets Public Views with UDCs (including Subsets in a Subset) 1st Time View activity following Cube Loading (Cube Dependencies) Best Practice - Separate Meta Data Loads from Data Loads to lessen lock duration Run TI command line 15. 2011 IBM CorporationInformation Management15BI Integration TM1 9.5.2 Conforms with C8.4 andCognos 10 BI Reporting on TM1 is faster BI Server memory not a bottleneck Zero Suppression reporting against large,sparse TM1 Databases having bigdimensions is much faster Top Count, Measure filtering, andAttribute filtering reports are much faster Reports indirectly referencing members vialevels or children of consolidation arefaster Drag n Drop TM1 iWidgets to BusinessInsight dashboards TM1 iWidgets adopt BUX toolbarGoal Cognos 10 Conformance Improve BI Report Performanceagainst TM1 Data Sources Leverage TM1 for suppressionand filter reports Optimize MDX generated by BI Reduce network data transfers Polish TM1 iWidgets for BusinessInsightBenefitsNotes: Supports Cognos 10 Dynamic Query mode BI Package Support for DMR / SAP BW 16. 2011 IBM CorporationInformation Management16Single Portal with Zero-footprint Web InterfaceRDBMS DW OLAP DataTM1 OLAPPowerPlaySAP BWESSBASEMS AnalysisServicesHeterogeneous Flat FilesCommon Metadata, Security, Integration Services, Query Engine, Automation, Process MgtCognos BI with TM1GoogleSearchCognosMobileCognosOfficeBI AnalysisFor ExcelReporting , Dashboarding, Ad-hoc Query, Analysis, Scorecarding, Event ManagementTM1 Web, Executive Viewer 17. 2011 IBM CorporationInformation Management17Cognos 10 Business Insight Dashboard with TM1 iWidget 18. 2011 IBM CorporationInformation Management18 19. 2011 IBM CorporationInformation Management19Excel Planning Templates What-If AnalysisWeb Reports BI Dashboards & Scorecards 20. 2011 IBM CorporationInformation Management20Cognos 10.1 FP1 on TM1 9.5.2: Moderate Sized TM1 Datasource ReportingOnly the first report execution time ismeasured as this more closely matches areal life scenario of a user running a reportonce in a given user session. 21. 2011 IBM CorporationInformation Management21Contributor Usability Planners having budget responsibility formultiple nodes (cost centers, projects, etc.)can update some/all nodes in a singleContributor Session Perform update while comparing nodes Data Spread across nodes Planners can submit all nodes of a singleconsolidation in one action Approvers can review / update asubmission of multiple nodes in onesession Workflow page presents a single hierarchyfor dual role Approvers/Contributors Text measures used for commentary canhave fixed column widths and height Consolidation data spreading when allchildren are null / zeroGoal Extend Contributor 9.5.x Multi-Node Editing Enhance Contributor Workflowpage layout and content Provide p

Click here to load reader

Embed Size (px)
Recommended