The mainframe is dying: Long live the mainframe application!

Much has been made of the perilous future CIOs whose organizations rely on mainframes may soon have to navigate, but the reality of mainframe infrastructure’s long-term outlook is a little more nuanced than that, as two recent news announcements attest.

Fujitsu recently divulged it will end sales of its mainframes by April 2031, discontinuing support five years after that. But CIOs running workloads on Fujitsu’s GS21 mainframe family don’t need to cast around for a migration path just yet.

Fujitsu remains very much interested in the mainframe market, with a new model still on its roadmap for 2024, and a move under way to “shift its mainframes and UNIX servers to the cloud, gradually enhancing its existing business systems to optimize the experience for its end-users.” Those enhancements will include features such as predictive failure detection and business continuity support.

Coincidentally, the same day that Fujitsu gave notice it would discontinue sales of the GS21, IBM unveiled plans to deliver its IBM Z mainframe platform as a service on IBM Cloud.

Meanwhile, IBM has yet to announce any end to sales or support of its Z Series machines, and like Fujitsu, it too continues to develop its mainframe product line. In the company’s January 2022 earnings call, CEO Arvind Krishna said the next generation of mainframes, based on a new chip called Telum, will be released by July 2022.

When will Z end?

According to IBM’s official mainframe life cycle history, each generation of hardware typically remains on sale for 4.1 years, with an additional 7.4 years of follow-on service. Some have lasted longer, such as the first machine to carry the Z name, the z900 G1, which remained on sale for 5.5 years after its launch in June 2006. The z900 also had one of the longest periods of follow-on service, at 8.5 years, a record it shares with the z114, launched in September 2011 and destined to be supported through December 2022.

IBM’s current mainframe models, the z15 T01 and z15 T02, were introduced in September 2019 and May 2020 respectively, and the company still offers follow-on service for machines right back to the zEC12 released in September 2012.

All that means that, even if the next generation of IBM Z were to be the last (an unlikely prospect given its investment in the new Telum chips), the company would likely still be selling mainframes through July 2025 and supporting them through December 2033.

Z as a service

Unlike Fujitsu’s move, IBM’s initial Z-as-a-service offering is not about managing customers off their mainframes and into the cloud. Instead, it’s targeting test and development functions, with the goal of making it easier for enterprises to set up such environments whenever they need them, without having to leave costly excess mainframe capacity sitting idle the rest of the time.

Called Wazi-aaS, the on-demand offering will enable developers to launch a test or development environment in IBM’s virtual private cloud, based on stock images or images built from their own on-premises LPAR (logical partition) and running pre-installed or customized software.

Wazi is a suite of tools IBM introduced back in 2020 with which developers can write z/OS applications and then test them in a z/OS sandbox on their own x86 hardware. It enables programmers new to mainframe development to continue to use their IDE of choice, removing one of the barriers to assembling a software modernization team. Wazi supports IDEs such as Microsoft VS Code, Eclipse, Red Hat CodeReady Workspaces, or in-browser OpenShift workspaces.

IBM claims it can have a z/OS system up and running in 5 minutes with Wazi-aaS, and can run applications with 8 to 15 times the performance of an x86 sandbox. Most of us will have to take their word for it for now: Wazi-aas is available only through a private beta test IBM described as a “closed experimental” deployment.

In addition to this peek at Wazi-aas, IBM also recently released the IBM Z and Cloud Modernization Stack, a suite of tools for analyzing mainframe applications, creating and deploying APIs, and developing for them in the cloud, with tools to automate the deployment and management of applications on z/OS.

The future of mainframe-as-a-service

Fujitsu’s and IBM’s plans to offer mainframe-as-a-service are interesting for a number of reasons.

Renting computing capacity, whether it’s a mainframe or a rack of x86 servers, rather than paying to build it can help IT departments match infrastructure costs to business needs. That’s great for the test and development environments IBM is initially targeting — but also useful for production environments in an increasingly unpredictable world where enterprises can experience transitions from rapid growth to near-shutdown and back again in the space of a year.

Perhaps more importantly, putting that computing capacity in the cloud makes it easier to modernize legacy applications by rewriting or migrating parts of them to other cloud infrastructure, or by augmenting them with cloud-based AI functionality. While Fujitsu is likely to use its own data centers for any hosted mainframe offering, it already has close relationships with Amazon Web Services and Microsoft Azure to deliver automated mainframe modernization services. IBM, meanwhile, offers a range of AI tools in its own cloud that developers can integrate into their applications, mainframe or not, to create chatbots, perform natural language processing, and automate workflows.

Career counseling

Fujitsu’s promise of 13 more years of mainframe support, and of an afterlife for those applications in the cloud, and IBM’s commitment to providing mainframe modernization tools, are good news for CIOs looking to protect their existing investments.

But just as important for many enterprises, said Pund-IT principal analyst Charles King, is finding, hiring, training, and keeping talented developers, and IBM’s new solutions are clearly focused on addressing that.