Abstract: Strategic and mission-critical architecture themes in North-Europe's largest IT-project conglomerate.
The bank is replacing hundreds of old applications by new off-the-shelf platforms in several parallel, long-running step-by-step programs. Consequently, many applications will be in transition for years. Adding up all the moving parts yields an extremely complex picture, and the devices for dealing with it emphasize what architecture is all about. In effect, the target architecture is irrelevant, what matters is a flexible "transition" architecture; the transition architecture IS the target architecture. An obvious idea is to hide the old/new distinction between an interface, but this is not feasible for several reasons. So what to do instead? I'll adress that question and also say a little bit about enterprise architecture modelling - information architecture, business architecture, integrations, dependencies and dynamic versus static modeling.
SOA and Microservices
The bank introduced a service-oriented architecture in 2008. Now, 10 years later, all the lessons have been learned, good and bad. The bank has recently introduced a microservice architecture, which implicitly is supposed to overcome difficulties experienced in SOA. But if care is not taken, some major challenges may be left unadressed, and new serious problems could be introduced, which are as fundamental as those of SOA. I'll discuss the relationship between the two paradigms in the bank, the challenges and promises of both, in particular what has been the main problems of the past, and how microservices can help. I expect the audience will have strong opinions in this area.
The bank has recently built a Big Data platform around the Hadoop eco system, including Kafka, and an event-driven architecture has been built around kafka. Another slightly earlier event-driven architecture is built around Oracle's SOA Suite, JMS and MQ. Many legacy applications in the bank have traditionally exchanged end-of-day files, but this approach may lead to inconsistencies between different ways of presenting data to the customer. With event-driven architecture this limitation can be overcome and new possibilities arise for the bank's applications to provide a much improved user experience. The question may be how far the idea should be taken - good bye to all end-of-day or synchronous dependencies, as some recommend?
As mentioned above, the bank has built a Hadoop eco system. This platform offers many new aopportunities. It could be the basis for a new Data warehouse infrastructure, it could be a gold mine for analytics, and it could be a place to get the 360 view of the customers' data for operational use. I'll discuss some of the challenges as well as some of the enourmous benefits that can be harvested here, and hope the audience can contribute their ideas too.
Morten Heine Sørensen
Senior architect with expert knowledge on Java EE and Oracle. Specialized within Performance optimization and development of commercial critical and heavy applications. More than 15 years of experience from rolls as architect, team lead and developers. Great knowledge within the finance and logistic industries. Focus on deliverance of durable solutions, that are building bridge between business and IT for the customer.