I’ve been getting a steady trickle of inquires this year about the future of the mainframe from our enterprise clients. Most of them are more or less in the form of “I have a lot of stuff running on mainframes. Is this a viable platform for the next decade or is IBM going to abandon them.” I think the answer is that the platform is secure, and in the majority of cases the large business-critical workloads that are currently on the mainframe probably should remain on the mainframes. In the interests of transparency I’ve tried to lay out my reasoning below so that you can see if it applies to your own situation.
How Big is the Mainframe LOB?
It's hard to get exact figures for the mainframe contributions to IBM's STG (System & Technology Group) total revenues, but the data they have shared shows that their mainframe revenues seem to have recovered from the declines of previous quarters and at worst flattened. Because the business is inherently somewhat cyclical, I would expect that the next cycle of mainframes, rumored to be arriving next year, should give them a boost similar to the last major cycle, allowing them to show positive revenues next year.
My crude and conservative guesstimate for IBM's mainframe hardware and software revenue is somewhere around $3 – 4 Billion, and based on the nearly constant litany of customer comments about pricing, very profitable. Adding in service revenues that the mainframes pull in that are not accounted for in STG revenues, and the overall mainframe business is probably in excess of $5 Billion, and probably the most profitable portion of their STG revenue, which also includes their Power RISC servers, storage and networking. Financially-minded readers should note that 2014 STG revenues will also include most of the year's x86 server revenues, but as of October this line was transferred to Lenovo, reducing their total revenues but almost certainly increasing their margins. So in terms of its underlying economics, IBM's mainframe business is inherently attractive to them, and my opinion is that they will not by choice abandon it for the foreseeable future and will continue a healthy investment in it, some of which is also leveraged by their Power RISC systems as well.
Customer Workloads and Intentions – The Dominant Variable
The question about the role of the mainframe then devolves to the underlying motivations of mainframe users – will they stay or will they migrate to nominally lower-cost platforms? I think the answer is kind of a blended analysis against a rapidly changing technology and workload background. While many workload have indeed migrated, primarily to RISC Unix and to a lesser extent to x86 Linux, much of the mainframe workload is still anchored to the mainframe by two underlying issues – software and overall scaleability and reliability. If you have a workload dominated by what we can call, to simplify our analysis, a "legacy" workload dominated by COBOL, CICS, and database (mostly DB2, maybe some IMS), migration from the mainframe is difficult and fraught with project risk, and the inevitable changes to overall DR/HA architecture almost always end up pushing the end-state cost of the new environment much higher than proponents expected in the early stages. In regards to scale, the mainframe is still the highest performing platform for OLTP, the undisputed king of batch, and the only platform that can gracefully handle mixed workloads – in short all the things that the mainframe has always done well.
At the same time, IBM has done an excellent job of enabling the mainframe as a platform for new workloads, notably Java and Linux, with special pricing for Linux/Java IFLs that place the overall life cycle costs much closer to x86 costs than ever before. While I remain a bit skeptical of these models until I can review them further, the overall growth of Linux workloads is impressive, and probably accounts for much of the growth in mainframe MIPS over the last few years. The growth seems to be concentrated in applications where access to mainframe-resident data is at the core of the application, environments where the reduced latency and high throughput that can be obtained by a Linux running on an LPAR on the same system as the database resources compensates for any cost differences. I expect that future mainframe product cycles will continue to push the mainframe as a hub for application services that depend on mainframe resident data even if the services are implemented in Linux/Java and other runtimes usually associated with distributed x86 systems.
Wrapping it Up
As I noted above, this train of analysis leads pretty directly to the conclusion that mainframes, while probably never breaking out of at best a low single-digit growth, will not go away for the foreseeable future, and that unless you are willing to engage in radical, risky and expensive transformation, your current large mainframe workloads will be with you for the long-term.