How to Survive the New Normal of Mainframe

How to Survive the New Normal of Mainframe Craig Hodgins 1

The World Has Gone Distributed But…. 80% of critical data is stored on mainframes Billions of transactions are executed daily on mainframes “If every Mach system failed, it would be front page news in the New York Times. But if every MVS system failed, the New York Times would not publish. And if every MVS system failed for a week, the New York Times would never publish again. Western civilization would fall because western civilization runs on MVS. ” Bob Rogers, Distinguished Engineer, IBM 2

The New Normal of Mainframe Applications: Customer-facing Transactions: Increasing Applications: More Complex Workforce: Transitioning 3

Applications Are Customer-Facing 4

The Problem A slow application is considered a down application - we all seem to have ADD nowadays - see The Shallows by Nicholas Carr - see The Brain Rules by John Medina - the 3 second rule Customers can vote with their feet (or mouse clicks) Outages cost money and make for bad PR 5

The (Potential) Problem is Ubiquitous Amazon Banks e. Bay Retailers Government Stock Exchanges 6

Obamacare crashes on Oct 1 2013 The health insurance exchanges created under the Affordable Care Act opened for business today, and people have started hitting the websites. Only one problem: The Obamacare websites weren't working. As of 9: 30 a. m. , Healthcare. gov informed us: "The System is down at the moment. We're working to resolve the issue as soon as possible. Please try again later. " 7

Transactions are Increasing 8

Transactions Are Increasing – Why? More people use computers today More people use mobile devices to access their data today 9

Applications Are More Complex 10

11

Transitioning Workforce COBOL Green Screen Record albums Java GUI i. Tunes 12

Boomer Statistics 77 million Boomers in the USA (35% of the population) 8. 6 million Boomers in Canada July 1 2012 the median age in Canada was 40 1 st Boomer turned 65 on Jan 1 2011 Kathleen Casey-Kirschling, born one second after midnight in Philadelpia on Jan 1 1946 Someone turns 50 every 7 seconds For the next 20 years 8000 people a day will turn 65 13

The Cost of Poor Performance Detail Design Programming System Test “The later performance problems are caught in the life cycle, the more costly they are to fix. Inefficiencies introduced in design can cost twice as much to fix during programming, four times more during system testing, and eight times more when the application enters production. ” Accenture User Test Production 2 x 4 x Relative Cost to 8 x Fix 15

The Cost of Poor Performance - War Rooms Can involve 20+ people Can take hours to days Finger pointing and defensiveness Lost productivity 16

Additional Costs of Poor Performance • Lost revenue • Penalties in SLA’s If you don’t monitor your applications, your customers will. Monitoring done by customers is the most expensive kind there is. 17

The Answer • Need visibility into the application performance before it becomes critical • • • Need timely resolution (identify the root cause) Automated Intelligent Easy Integrated 18

We Need Something That… • automatically provides the answer in one place for faster and better root cause analysis and ROI 19 19

Txn took 295 ms Without the agent for z/OS, all activity on the Mainframe is a blackbox! 66% of Exec Time spent in call to Stored Procedure Compuware Confidential 20

Complete End-to-end Visibility From Mainframe to User Mainframe MIPS Growth on Mainframe ü Poorly performing mainframe code and SQL Compuware Confidential Servers Real Users Internet/Cloud Last Mile Transactions off Mainframe ü Java and. NET code calling CICS and MQ 21

Case Study: European Bank Off Mainframe MIPS Savings • Chatty distributed application driving CPU consumption. One business transaction invoked 51 CICS transactions • Overall CPU consumption reduced by 7. 5% through rearchitecting the application Annual savings = $250, 000 On Mainframe MIPS Savings • Bad performing SQL Statement was identified immediately • Average CPU consumption: what took 3 CPU seconds before was reduced to microseconds after tuning • 80, 000 executions per year • Associated CPU costs of $2, 500 per CPU hour Situation • Company extending e-banking services • Composite application including distributed and mainframe tiers • Through new application architecture mainframe transactions are growing significantly Results • Enabling teams to resolve performance problems before customers affected • Meeting requirements of IT auditors to monitor service quality • Reducing MTTR • All performance issues captured • Transaction details visible through all tiers, from end-user to the mainframe Annual savings = $170, 000 22 22

Transaction Flow Diagram 23 23

56. 69% of CPU at 0098 A CALL CEEGMT (LILIAN, SECONDS, FC); 24 24

Trend and Summary Reports 25 25

26

27

28

The APM Capability Maturity Model APM Practice Commitment Level 5 Kaizen Level 4 Disciplined Level 3 Process oriented Level 2 Proactive Level 1 Chaos Continually evaluate and improve the APM program Establish accountability for application performance. Track, measure and report on the APM program Define processes for performance evaluation at established checkpoints Reclaim production resources through repeatable projects Respond to production crises Organizational Benefits 29

APM Capability Benefits APM Practice Commitment Level 5 Kaizen Level 4 Disciplined Level 3 Process oriented Level 2 Proactive Level 1 Chaos Maintain competitive edge Reduce cost of achieving application efficiency Prevent deployment of inefficient applications Defer upgrades and improve service levels Resolve production crises Organizational Benefits 30

A Final Thought…. Computers are useless. They can only give you answers. Pablo Picasso 31

• Questions? • Comments? Thank You 32