I started working on the mainframe less than two years ago and immediately recognized that System z offered a technically superior platform for deploying Linux when compared with many distributed systems. Nearly two years later the deployment level of Linux on the mainframe has remained largely static; certainly we have not seen the growth that you’d expect of a technology that underpins a high-growth space such as the cloud.
As we struggled to reconcile the anecdotal evidence from our customers with the information posited by vendors in the space, CA recently embarked on a concentrated effort to examine our customers’ use (or lack thereof) of Linux on System z.
We decided to do some product management 101 fact-finding, consisting of a number of online polls and interviews with advocates, opponents and moderates. We wanted to determine when customers elected to use System z as a platform, the compelling events behind those decisions and roadblocks to more deployments. The results are incorporated in Figure 1.
We mapped the typical Linux on System z workload on a graph showing the internal “resistance” to deploying Linux on System z against the size of the addressable market. We found five broad workload categories, which we mapped to the chart, discussed below.
Supporting Internal Mainframe Systems
In the first column we see Linux workloads that exist to offload billable MIPs from z/OS. In all cases, there was little resistance from the enterprise IT group because they typically were not even aware that these workloads existed. Even when they were informed, they could not make a coherent argument that these workloads should be moved off the platform.
Typically, these guests are very limited in number and complexity. They are not supporting high-growth areas such as the cloud so they will scale only as General Processor (GP) MIPs scale, and that’s painfully slow.
Benefiting From Proximity to z/OS
This category was very similar to the first one but with independent applications that make heavy use of a z/OS system and benefit from a direct connection. In theory, you could deploy these workloads off the platform but they are so mainframe-centric that the work involved in moving them would discourage most enterprise operations teams. Once again, though, these workloads are almost entirely tied to the scale of the underlying/OS systems and are, hence, small and unlikely to grow at any speed.