It all began on Dec. 18, 1999, whenIBM released a collection of patches and additions to Linux kernel version 2.2.3. This event started the relationship between the Penguin (aka Linux) and the mainframe, and was essentially the beginning of the System/390 kernel for Linux on System z. Then there was the official public announcement of Linux on System z and Integrated Facilities for Linux (IFL) in 2000, along with a whole ecosphere of software and tools to support Linux on System z produced by IBM and IBM partners. Many large and small enterprises have saved millions of dollars by adopting Linux on System z in the past 15 years.
Trouble in Paradise
Yet with mere months until the Penguin and mainframe celebrate their 15th anniversary, the relationship may not be where it should be. Let’s consider some facts. According to multiple surveys, 53 percent of mainframe customers own at least one IFL but only 41 percent of IFL owners are actually using them. Linux on System z adoption is low despite what might be expected of a 15-year-old technology with the marketing might of an IT industry leader behind it. In an effort to better understand the dynamics of the Linux on System z market, a series of in-depth discussions were conducted with Linux on System z advocates, opponents and everyone in between. From the tales of phenomenal success as well as horrific failure, four themes emerged.
“Immature and Unproven”
The most common theme was that Linux on System z is immature and unproven. This theme was very popular among those who had never tried Linux on System z. The comments here reflected the other three themes. In corporate IT, there seems to be a lack of knowledge about mainframes as well as the pricing model of Linux on System z, plus the fact that operating Linux on System z is an exception to business as usual. The remaining three themes, discussed below, help contribute to the perception that Linux on System z is immature and unproven.
The statement that the mainframe is misunderstood was mentioned in all but one or two interviews. Even at enterprises where the Penguin and the mainframe have been highly successful, there was a perception that the mainframe is misunderstood and causes friction between those who run the distributed platforms and those who run the mainframe. In some cases the interviewees believed that distributed teams see the mainframe as a black box with no transparency, even as a threat to the job security of distributed teams.
There has always been agreement that implementing the Penguin on the mainframe is not inexpensive. While the cost to run workloads in volume is cheaper in the long run, the cost of an IFL and all associated software in a single chunk is too high compared with the much smaller incremental costs of adding distributed servers. As a result, when some enterprises require an additional IFL, they opt to go the distributed route. It’s a lot easier to digest and justify a $10K expense than a $200K expense.