System Z Performance & Capacity Management using TDSz and DB2 Analytics Accelerator: UnipolSai Customer Experience Marina Balboni & Roberta Barnabé System Z Transactions and Data Area, UnipolSai Francesco Borrello Technical Sales and Solutions IBM Sales & Distribution, Software Sales Milan, 8th April 2014 Rome, 9th April 2014
Agenda 1 Customer Environment Customer needs Pain points 3 IBM Proposal: Capacity Management Analytics UnipolSai implementation 5 4 Final results Next steps 2 2 6
Agenda 1 Customer Environment 2 3 3 4 5 6
Who is UnipolSai Second Insurance Group in Italy, first in Non-Life business Total premiums of 15.4 billions About 15000 employees Company with the biggest number of branches in Italy 4
UnipolSai Hardware Technical Environment 2 IBM 2827-H20 (707 + 704) 707 Production & Development (ex Unipol) : o 7 GP + 3 ziip + 6 zaap + 1 ICF o 512 GB Memory o 1092 MSU 8954 MIPS 704 Production (ex Fonsai): o 4 GP + 1 ziip + 1 zaap + 1 ICF o 512 GB Memory o 664 MSU 5409 MIPS 14363 MIPS 1756 MSU 5
UnipolSai Hardware Technical Environment Appliance: o IBM DB2 Analytics Accelerator for z/os: N1001-05 Storage MGM Configuration o DS8870 DS8800 MM (Sync second site) o DS8800 DS8700 GM (Async third site) o 2 X (TS7720, TS7680) Virtual Tape + TS3500 Real Tape 6
UnipolSai Hardware Technical Environment Disaster Recovery Site IBM 2817- M15 o 7 GP processors o 1 ziip SE o 5 zaap SEs o 1 ICF o 165 GB Memory Three sites configuration 7
UnipolSai Mainframe Technical Environment z/os version 1.12 DB2 version 10 NFM (Data Sharing Implementation on going) 9 Subsystems IBM DB2 Analytics Accelerator V3.1 (Migration to V4 on going) CICS TS 4.2 50 Subsystems 9 million transactions/daily WAS 8.5.5 + WAS 8.5.5 on z/os 11 Application Servers (2 clustered) 6.5 million threads/daily WebSphere MQ 7.0.1 8
UnipolSai Application Environment 9 Cobol Cics/Batch Static and Dynamic Assembler JAVA (SQLJ/JDBC) DELPHI (ODBC) on Workstation Visual Basic -.NET on Workstation
Agenda Customer needs Pain points 2 1 10 3 4 5 6
Customer Needs Pain Points SMF is preferred source of statistics information helping System Z capacity management, but: Lack of a unique tool for collecting SMF data Several monitors with overlapping functionalities Poor quality Redundancy High complexity in correlating data Lack of system resources for collecting detailed information DB2 for zos V10 gives some relief allowing SMF data compression Need of saving huge amount of historical SMF data Storage issue Performance and trend analysis 11
Customer Needs Pain Points Solution should follow cost reduction directive, so the consumption of MIPS and use of storage should be reduced and kept as small as possible Complex analysis show elapsed time and cpu consumption issues Improvement in the existing user interface should be provided, allowing to: Intuitive navigation and easy-to-use tools Scheduling and ease of distribution Different report output formats 12
Agenda IBM Proposal: Capacity Management Analytics 3 1 13 2 4 5 6
Capacity Management Analytics Core Architecture 14
IBM Cognos BI Cognos Business Intelligence provides the range of analysis capabilities necessary for optimizing zenterprise use by confidently and simply compiling the information necessary to understand and manage system activity while significantly improving the ability to identify potential issues and pinpoint their cause. 15
IBM SPSS Modeler SPSS Modeler can help you use predictive analytics to forecast future requirements for zenterprise and ensure the capacity required is available when the business needs it 16
IBM Capacity Management Analytics Solution Kit Pre-built interactive reports and models Several COGNOS reports (CPU, WLM, Memory) SPSS predictive analytics model that forecast LPAR CPU usage 17
IBM DB2 Analytics Accelerator: Faster Analysis IBM DB2 Analytics Accelerator What is it? A high performance appliance that speeds analysis, enabling you to base your projections on a larger sample of historical data What does it do? Base forecasts off larger samples of historical SMF data to improve accuracy of predictive models Dramatically accelerate the analysis of your zenterprise usage & performance data Significantly speed up complex queries of the large volumes of data that are being created by zenterprise. Lower the cost of long-term storage of large volumes of historical SMF data with a highperformance storage saver feature 18
IBM DB2 Analytics Accelerator: HPSS Reducing the cost of high speed storage Time-partitioned tables where: only the recent partitions are used in a transactional context (frequent data changes, short running queries) the entire table is used for analytics (data intensive, complex queries). DB2 partitions are deleted after the High Performance Storage Saver are created on the accelerator DB2 #1 Query from Application No longer present on DB2 Storage Or 19 Accelerator Accelerator Accelerator Accelerator Accelerator Accelerator Accelerator #1 #2 #3 #4 #5 #6 #7
Agenda UnipolSai implementation 4 1 20 2 3 5 6
Overview of UnipolSai Solution Architecture SMF logs TDSz Most recent data (7-30 days) + NNN weeks stored in HPSS Cognos DB2 Reports 21 Most recent data (7-30 days) <= 1TB
The Plan to Implement the Designed Solution Phase 1: Setup & Configuration Setup of TDSz and Cognos environments Phase 2: Test TDSz and Accelerator Synergy Evaluate IBM DB2 Analytics Accelerator V3.1 functionalities Evaluate performance and consumption benefits Phase 3: Speed up and Archive Tables partitioning and archiving Batch alignment of data on the Accelerator Force queries to execute on the Accelerator 22 Phase 4: Implement a complete automated process
Phase 1: Setup & Configuration SMF record types collected via TDSz z/os, DB2, CICS, WAS, MQ Configured both tables with detailed data (timestamp or hour) and tables with aggregated measures COGNOS: setup and connection to TDSz DB2 for zos database Migration of existing reports and definition of new ones Installation of TDSz provided reports Migration of existing reports to COGNOS format with graphs and/or columnar data Development of new local reports to satisfy new requirements Reports scheduling and automatic distribution inside UnipolSai Schedule of predefined reports with output in different formats, mainly PDF Automatic distribution of reports to defined users Automatic publication of reports in Intranet Website 23
Phase 2: Test TDSz and Accelarator Synergy Evaluate IBM DB2 Analytics Accelerator V3.1 basic functionalities in a Test Environment Add and load some tables on the Accelerator Test sample queries execution on the Accelerator Evaluate TDSz and IBM DB2 Analytics Accelerator synergy Identification of a Test Environment with a few TDSz data Identification of some TDSz critical reports and queries Add tables, load data on the Accelerator and execute test cases Evaluate performance and consumption benefits in Production Environment Identification of the most critical queries in terms of elapsed time and MIPS consumption to be used for benchmarking Forcing execution on IBM DB2 Analytics Accelerator SET CURRENT QUERY ACCELERATION ELIGIBLE; Analyze results From almost 4 hours to 1 second 24
Phase 3: Speed up and Archive Identification of TDSz tables for partitioning Efficient loading and archiving purposes Choice of partitioning criteria (time criteria) Partitioning by month for tables with detailed data Partitioning by year for tables with aggregated data Partition TDSz tables Drop and re-create Archive tables (on going) Development of REXX program for batch data alignment on the Accelerator Only modified partitions or tables 25
Phase 3: Speed up and Archive 26 Configure Cognos for forcing both reports development and reports execution on the Accelerator
Phase 4: What about a complete automated process? Implementing a complete automated process which consists of the following steps orchestrated by Tivoli Workload Scheduler (TWS): Disable TDSz accelerated tables Mandatory only if QUERY_ACCELERATION zparm is different from NONE Start TDSz tasks for collecting SMF data and saving extracted information on DB2 for zos Invoke REXX batch program for loading only modified TDSz partitions or tables after collecting operations Enable TDSz tables on the Accelerator Trigger Cognos reports execution and distribution 27
Agenda Final results 5 1 28 2 3 4 6
Cognos Reports: Measured Elapsed Time 29
Cognos Reports: Measured CPU Time CPU consumption comparison between the two following scenarios: Scenario 1: all reports queries execute against DB2 Scenario 2: all reports queries execute against IBM DB2 Analytics Accelerator 30
Some Numbers about Space Usage PDT_OBJECT_150 is not included in the graph, as it's too big and it's out of scale 31
Load Operations: Measured Elapsed and CPU Time 3 Jobs Output: End of month elapsed and cpu time 32
Ease of Sharing within UnipolSai 33
Ease of Sharing within UnipolSai: Report Visualization 34
Just Another Example of Cognos Report Output 35
Agenda Next steps 6 1 36 2 3 4 5
Next Steps: Ideas for the future we're working on Use only TDSz tables containing detailed data Avoid to update TDSz tables containing aggregated measures All the aggregated measures will be calculated on the fly by DB2 Analytics Accelerator Use of SPSS to forecast resource requests SPSS can forecast future capacity to ensure the capacity is available to satisfy business needs Use of IBM DB2 Analytics Accelerator V4 (Migration on going) Performance improvements Archive operations improvements (automation of previous archive operations) Static SQL 37
Acknowledgements Fabio Riva, zchampion, zstack Advocate, zclient Architect, IBM Italy Fabio is a Senior IT Architect in SWG IBM Italy. Joining IBM in 1985, he covered different positions inside the company, from MVS SysProg to Systems Engineer, up to Senior IT Architect. During the last years his main activities were related to business development on System z. Fabio followed the development of cross-platform (hybrid) solutions, having the main focus on mainframe platform. He's also supporting some Tivoli products in the areas of SW asset, licence, and cost management on System z. Fabio published several publications inside IBM, but also articles on external newspapers. He acted also as a speaker at several international conferences and technical user groups. Francesco La Sala Senior Consultant - 5EMME INFORMATICA S.p.A Francesco is a Senior Consultant in 5EMME INFORMATICA S.p.A. and he s providing a great contribution to CMA project with UNIPOLSAI customer. His areas of specialization are in Tivoli (now CS&I) software brand, in particular regarding the product TDSz (Tivoli Decision Support for z/os). He has deeper knowledge also in DB2 area, in particular on IBM DB2 Analytics Accelerator appliance. 38
Thank you for your attention! 39