"Everything should be made as simple as possible, but no simpler" - Albert Einstein
Once existed a world where all software was FREE, not chargeable to customers and only hardware contributed to sales and revenue. Too many customers wrote custom software on their hardware causing interoperability issues which led IBM to create System/360. This created the era of super computers called 'Mainframes'. When they came and conquered, Mainframes were considered almost super natural - modern, robust, multi programming, reliable, secure and with high throughput.
Naturally there were lot of adoptions leading to monopoly and a 'King' was crowned. Most of the financial corporations acquired them and marketed security and performance of Mainframes to gain customer confidence. The systems lived to the hype and delivered the promise. With no alternatives to challenge, the hold that Mainframes had on transaction processing only became tighter and locked in a flock of customers.
Time went by and slowly, just like the anti-biotic resisting bacteria, the normal computers of the world grew stronger and were available cheap. The gap is becoming shorter and shorter with scale-out, redundancy with cheaper hardware and increasing processing and memory power. The 'closed source' strategy of Mainframe has ensured the vendor lock-in and at the same time, has created complexity around sizing and capacity planning.
Today,not all is well in the King's world. The technical debt built by the monolithic spaghetti code over the time has become the nightmare of CTO's to manage or think about re-writing. The difference in the way back-office operations are run now in financial world also had demanded more real time, analytical abilities out of the back-end system. The business demands customer-centric flexible and agile systems to compete with new comers like FinTech companies.
With new business goals and technical abilities of commodity hardware, now is the 'Perfect Storm' that is brewing to try and dismantle the crown. The modern distributed systems are definitely showing equivalent, if not better, alternatives. For example, Apache Spark can be an alternative for batch workload processing outside Mainframe, providing almost same and sometimes better scalability and throughput.
Even though the alternatives are emerging, Mainframes are still a better and default choice for complex workloads like transaction processing of structured data, query processing on RDBMS back end, etc. especially when the volume is high and robust throughput is a demand. This is true for most of the financial organizations' core functions that act as bread and butter of their business.
Infosys approach after weighing in all this is 'True-Core and Core-Surround' so as to paraphrase Albert Einstein's quote at the first line, "Only the Core functionality need remain on the Mainframe, nothing more". The core-surround can work in simpler environments where Mainframe is not called for, whereas the true-core is reserved for the platforms that can provide the complexity that is there.
Consider a banking organization where this approach can be extended to categorize business functions as below:
True-Core: Core functions should create and manage accounts, record transactions, maintain ledger information and provide account related information to all systems within the bank across products lines. Calculation and posting of interest is also a basic accounting function performed on an end-of-day basis within core system. This function can be retained in the Mainframe system.
Core-Surround: Mortgages should be managed through a product based platform to segregate the processes and channels managing the mortgage through its life cycle. This functionality can be moved out of core banking system.
This simplification of core leads to functional consolidation, faster implementation, process optimization, complexity reduction, Microservices based Architecture etc.
Extending the strategy our approach is to have a top-down view of related business functions and categorize them as System of Innovation, System of Differentiation and System of Records (as per Gartner Pace Layered Architecture). Normally the System of Innovation and System of Differentiation are good candidates for core-surround, and the System of Records for true-core. Once that demarcation and enabling of the True Core and Core -Surround has happened we can look at gradually even phase wise transformation of the Core. This approach follows the 'Strangler Application' pattern (Martin Fowler) that enables in transforming legacy. This leaves behind only what should essentially remain on Mainframe and nothing more.
Infosys "Transform" Model for Modernization
This pattern is something which has resonated with various clients that we have talked to and also which we are helping to implement.
For example, for a major European Bank we are helping on transforming their monolithic Core Accounting system, which is a very tightly coupled system with multiple functionalities incorporated into one humongous Mainframe application. We are doing this by looking at the various Technical and Business Functionalities to help come up with a multi-year road map to phase wise move to a Core and Core-Surround model to bring better efficiencies.
For a cards client we have evaluated their loyalty platform, as they were tied down by platform limitations leading to missed business opportunities on time. We followed a 'meet in the middle' approach analyzing the current business process and the systems in place. A 1-3-5 Year road map was established and the journey started. The end result is expected to be an agile system with around 30% reduction in cost and 30-40% improvement in time to market.
Nareshkumar Manoharan and Tiju Francis