4 83 54 2 The SDC Mainframe migration project 20 th September, 2012 Robert Elgaard, re@sdc.dk
About Corebank The Modernization project is targeting the entire Corebank application (customers, accounts, transactions, interest calculations etc.) covering alltraditional banking products. SDC Corebank consists of: 14,500 Cobol modules 12,1 mill. lines of code +2,500 online transactions +1,200 unique batch jobs (expands to +30,000) +49,000 function points (http://www.ifpug.org/)
Motivation The only motivation for migrating Corebank from Mainframe to Microsoft.NET was a need to cut operational expenses. Using in excess of 9500 MIPS, the MIPS payments became a burden for our ability to compete in the market. The project was initiated as a part of a comprehensive strategic cost efficiency program that we run at SDC. Microsoft gained interest in the project, and eventually it became a part of Microsoft Mission Critical Program. From autumn 2007 to early spring 2009 a comprehensive pre-study was conducted. The project was launched in May 2009.
Other benefits Access to a modern development environment like Visual Studio Skill shortage - people with mainframe skills are getting harder to employ More agile and flexible infrastructure Strongly improved procedures and disciplines for server operation Improved monitoring capabilities by health modeling and SCOM Economy of scale A stringent, streamlined and future proof architecture
A simple rehosting concept as migration strategy Development General mainframe migration strategies: Check-in AppBuilder -Replace Nebula (IDE) (Source repository) -Rewrite Check-out -Rehost Build & Deploy pre-process Build & Deploy Mainframe - Two parallel runtime environments - Single source base - Supporting two technologies - Supporting 138 banks - Supporting four countries Microsoft.NET - Minimal impact on developers -Migration without any stop in development
Implementation phases Additional 4 mill. Euro in benefit per year. Benefit 4 mill. Euro per year. Additional 7 mill. Euro in benefit per year.! Plan lacked and analysis failed May 2010 May 2011 February 2012 May 2012 Phase-1: 70 query transactions and first version of the operation platform Phase-2: All transactions (employee portal), some netbank functionality, batch and full operation platform Phase-3: remaining netbank functionality and migrate DB2 to SQL Server. The entire project has been broken down into these three independent phases, each phase having its own risk-, cost-and benefit profile.
Implementation first phase AsWas AsIs Emulate Corebank communication progressive implementation ziip
implementation second and third phase ziip Shadow Master
Implementing batch Mainframe Batch Central OPC.NET Batch jobs ziip During the transition process there will be mechanisms in place to facilitate instant fallback to the existing batch job on the mainframe. At the end scheduling of batch jobs are moved to the.net platform, either using a traditional scheduler or by implementing workflow based batch procedures.
Business case Assumptions for business case calculation: Financial benefit is calculated relative to the as-is situation, where we stay on mainframe until end of current IBM contract, which expires 31. marts 2017. From experience we know that MIPS consumption grows between 10% and 15% each year. The benefit is fully realized at the end of the IBM contract, and the calculation did assume 10% as yearly MIPS grow. The calculation includes reduction for licenses paid to other mainframe software vendors who base their revenue at MIPS. This includes CA, Compuware, BMC and more. SDC has chosen to let a large part of the MIPS consumption be OnDemand.
Business Case -MI cost model MIPS consumption 12500 11500 10400 End 2017 = 18.500 MIPS Revised forecast (15,4 % growth per year) New conservative forecast 9500 9000 8600 8000 7400 5550 Actual MIPS consumption in 2010 Contractual baseline Contractual limitfor MI reduction 2008 2009 2010 2011 2012 Original forecast (10 % growth per year) OnDemand MIPS: 100 % saving on 10400 7400 = 3000 MI (2012 MI consumption according to baseline) 75 % saving on 7400-5550 = 1850 MI ( max- baseline reduction is 25%) 0 % saving for MI consumption below 5550 MI År
Reducing operational cost Significant operational cost savings: Mio. Euro 0.3 mill. 0.7 mill. 0.5 mill. 1.3 mill.. Yearly savings exceeds 18 mill. Euro 0.4 mill.. 22 (2014) 4,0 0.8 mill. Approx. 18% of current mainframe cost 3-years software subscription Hardware is deprecated over 3 years
Project cost The project cost accumulate to approx. 24 mill. Euro. The overall cost distribution is as follows: Hardware/Software : 4 mill. Euro Application migration : 11 mill. Euro Operation platform, procedures & processes : 9 mill. Euro Total yearly cost inclusive operations, software and deprecation approx. 4 mill. Euro ROI is approx. 2 years.
Production setup and economy of scale Production setup looks like this: Fujitsu-Siemens RX300 S5 Server, 3,33 GHz, 8 Core, 48 Gbyte memory Windows Server 2008 R2, Data Center Edition on VmWare ESXi 4.0 Price less than 10.000 Euro, plus 16.000 Euro per year in OpEx Power of the sizing unit corresponds to 2000 MIPS
Monitoring Significant funds has been invested in order to establish a mainframe class operations environment. The monitor and operations environment is based on Microsoft SCOM (System Center Operations Manager). The SCOM configuration has been derived from a health model. This model was defined by deep dive analysis of the most important system components.
Monitoring for stability All application components and system resources are monitored by SCOM. Performance counters are injected into application components to achieve detailed runtime information at module level (memory-, handles-, CPU utilization )
Problems encountered Peopleware Competences and responsibilities - Make responsibilities crystal clear. Collaboration between highly skilled people from different organizations was challenging. All parties tend to over-contributes to the solution. Sub optimization Lack of enterprise view. Vendors were occasionally over-focused on designing the Corebank application to perform well. On few occasions the Corebank.NET centric design focus has not been optimal for SDC s entire application portfolio. Understand that the migrated application is only one applicationamong hundreds of other applications. Technical HIS memory leak Cobol pointer limitation Floating point consistency Memory initialization for Cobol structures Legacy applications needed optimization since they were written for different topology Batch Performance SQL Server and DB2 differences Interfacing mainframe appplications Business relations In the beginning of the project it was, for obvious reasons, difficult to convince our facility manager IBM of the benefits of the project. When IBM realized that the migration was for real, they have been working - by the book, and without flexibility - with their tasks and project assignments.
Status All +2,500 online transactions are running on.net. All batch jobs are running on.net. All netbanks transactions are running on.net Banks will be migrated from mainframe DB2 to MS SQL Server during September 2012 31 st May were the following number of transaction processed on the platform: Branches 4.867.088 Netbanks 4.505.214 Total 9.372.302 The project will offload in excess of 10.000 MIPS Current financial achievement: + 7,5 mill Euro per year.
Key learning points #1: A crystal clear and a rock solid implementation strategy is mandatory. - Best practices - Parallel runtime environment with flexible transaction routing. #2: IBM is a clever opponent; they do understand how to protect their business. #3: Hardware is ready, powerful and mature. #4: The.NET product stack is comprehensive and mature. #5: Be prepared to optimize legacy applications. -Address this during startup; it is a candidate for a joint design session. #6: Cheaper is better. #7: Stay focused
Thank you for your attention! Questions?