Mainframes stand as a telling case in the story of survivor technologies. While it may lose ground to insurgent options, Big Iron is still alive and well as major financial institutions remain loyal to the technology that has served them so well. Today, 96 of the world's top 100 banks use IBM's System Z mainframes.
Some might argue that newer architectures cannot match the benefits of mainframe computing for core applications in terms of reliability, availability, security and the analytical speed required by financial services organizations.
This set of beliefs is partly what keeps mainframes going, but the enterprise architecture head at a Canadian bank argues that it is only a matter of time until this is no longer the case. The executive, who asked to remain anonymous, says:
The mainframe will eventually be replaced with new products and services. But it will take some time for this endeavor: the culture and paradigm shift is not easy.
Avoiding the cloud
The demise of the mainframe has been predicted on many occasions and the footprint of these machines will continue to shrink, says Craig Lowery, research director at Gartner. According to Lowery, there is no technical reason why a bank wouldn't be able to achieve the mainframe's guaranteed performance in the cloud. He says:
I take issue with the point that some people make that cloud is unsuitable for mission-critical applications at large financial institutions, because this is not true. The myth is still out there and while there are some instances where cloud would not be suitable, such as some algorithms that can’t be distributed, these are exceptions rather than the rule.
Organizations which resist a move to the cloud often do so because of the cost associated with rewriting applications for the cloud, as well as potential disruption and risk to the business. Lowery adds:
Reliability, security, availability can all be achieved in the cloud, but the only way that can be achieved is by knowing how to build and deploy cloud-native software properly. We believe you can be more secure in the cloud than in a private datacenter.
It may take a while, but financial services firms will have to address the reluctance to migrate to the cloud at some point because mainframes will become more and more expensive to maintain, according to Lowery. This may also buy incumbents some time in terms of skills availability as sector-specific expertise needed for cloud migrations is still in short supply. Lowery says:
Most banks would love to have a Netflix type of [cloud architecture] expertise but it's expensive to hire those people or to outsource that kind of work. It's one of the reasons why they are saying that it's best to stick to mainframes until these skills are more available and understanding of how to build cloud-native systems becomes more pervasive.
The startup approach
Cloud is the go-to development and deployment method used by financial services newcomers. Alongside cost-effectiveness, the cloud-only approach also enables startups to better respond to customer and regulatory demands, as well as focusing on their commercial offerings.
Nubank is one of Brazil’s best-funded fintechs, with over 3 million customers across its credit and banking products. By the time the company’s first code was written in 2013, designing a greenfield project for the cloud was an obvious architectural decision, according to the firm’s co-founder and chief technology officer Edward Wible. He says:
[Being cloud-based] allowed us to focus our limited resources on the things that would differentiate us from the competition rather than worrying about hardware.
Wible consulted with mainframe technology vendors at the time Nubank’s core systems were being developed. None of the options were cost-effective enough for an early stage startup, and all required extensive customization and adaptation to the local market. He says:
In the end, the decision to write our own core banking software targeting the cloud was the only realistic decision we could make.
Benchmarks are often misleading and it can be difficult to compare mainframe and cloud performance without considering the specific workload in question, but Wible argues that in Nubank’s case, cloud is still the best option. The executive points out:
While mainframe computers may seem superior machines to the commodity virtual instances available in the public cloud, we find that the flexibility and agility to "scale out" is usually more valuable to us than "scaling up", although cloud providers allow you to scale in both ways.
High availability is generally achieved through redundancy; with the ability to autoscale in response to dynamic peaks in demand, we believe our availability is more than good enough for our business needs.
The financial services computational workload for any given customer is generally modest, Wible explains - so often Nubank can achieve more throughput by doing more in parallel instead of processing each customer faster.
The startup also draws a distinction between production workloads such as powering the firm’s customer-facing mobile app and analytical workloads such as creating reports for regulators. The company also uses different database software and different underlying hardware to optimize performance for each type of workload.
When it comes to security, the CTO believes that it is possible to design secure and unsecured systems for mainframes and for the cloud - the computing substrate being by far from the most important factor to consider.
While open banking can be implemented on a mainframe architecture in the same way that it can be deployed in the cloud, Nubank’s Wible believes that the main driver to staying competitive for sector organizations is culture - with cloud being key to shifting old paradigms. He says:
When there are separate groups responsible for development and operations, incentives become misaligned, with developers wanting to change things as fast as possible, while operations teams want to maintain things as stable as possible.
When small teams are empowered to own development as well as deployment to production and subsequent maintenance, this misalignment can be mitigated, and cycle times can be reduced.
Designing for the cloud makes it easier to combine these responsibilities in small teams, which we believe is the key to maintaining the culture of a startup, even as we scale.
Even though technology decision-makers in financial services accept that cloud will become pervasive and even touch some of their core systems in the near future, mainframes will still play an important role in their strategies for years to come.
As well as cultural factors, mainframe computing is cost-effective for large businesses than rip-and-replace cloud migration projects.
Mainframes may indeed be more cost effective to operate at scale in terms of monthly cost per customer, though as Wible says, this efficiency may come at the cost of flexibility. Large organizations might put cloud migrations off for now by arguing that the model is not appropriate for them, but the consequences of leaving it until later may well come back to bite them in the future as disruptive newcomers eat into the traditional markets.