Later this year IBM plans to have put in place a licensing model that will provide a compelling financial incentive for IT organizations to deploy microservices based on containers on a new z14 mainframe platform.
Mike Desens, IBM vice president of offering management for z Systems and LinuxONE, says IBM is making use of a proprietary container engine it developed for the zOS operating system to run Docker container images. Because IBM charges mainframe customers based on peak usage on the monthly basis, Desens says customers will be able to take advantage of logical partitions (LPARs) on a mainframe to run microservices as a nominal cost, if not for free outright.
The licensing approach to running containers on a mainframe that IBM will employ is modeled after IBM’s Linux on a mainframe service. If peak usage on a mainframe is driven by, for example, an online transaction processing (OLTP) application, any additional unused capacity in a specific month can be allocated for free to other classes of workloads.
Given the amount of data that can be housed on a mainframe, Desens says IBM is betting that many developers will want to deploy microservices as close as possible to where the most amount of data is being generated in the enterprise. In fact, IBM claims that 68 percent of the world’s production workloads run on a mainframe at only 6 percent of the total IT cost.
IBM also claims a z14 can support up to 2 million Docker containers running on its mainframe operating systems. Much of that capability can be attributed to the fact that the latest z14 mainframe can be configured with up to 32TB of memory, which, along with other I/O enhancements, increases overall z14 performance by a factor of three over the z13 series.
Not every IT organization can afford the cost of acquiring a mainframe. But those that can derive an economic benefit in terms of being able to run multiple types of workloads for a specific cost that is unrivaled by other platforms. In fact, IBM contends that once all those classes of workloads are accounted for, the return on investment (ROI) with a mainframe becomes extremely compelling. At the same time, IBM is also starting to make mainframes available as a hosted cloud service to drive the total cost of ownership (TCO) for a mainframe even lower.
The battle between mainframes and distributed computing systems over hosting workloads has now been fought over the course of several decades. Microservices based on containers simply represent another front in the fight. But given the latency sensitivity of many of microservices-based applications, it may very well turn out that mainframes are better-suited for containers than initially thought. Better still, because those containers can now run almost anywhere, that historic debate between mainframes and distributed computing systems is simultaneously becoming a whole lot less relevant to all concerned.