Linux on mainframes – 10 years old already

It seems so long ago, back in 1999, when we were discussing the Y2K bug (that was going to bring the universe as we knew it to an end) and whether Linux on a laptop was ever going to scale to mainframe proportions. And, while at the time I said the Y2K bug was overhyped (but then we all said we said that afterwards!!), I was never overly sold on mainframe Linux – wondering what kind of strange hybrid creature it would be. Doh! Not only has mainframe Linux (Linux on System z) now celebrated 10 years of service, it is a hugely successful and growing part of the mainframe world.

In 1999 IBM announced its commitment to Linux and promised to make it available on all their platforms and also produce Linux versions of its software. This meant that IBM would be spending lots of money on Linux and needed it to be a success. I can remember talking to various people at conferences, and the general opinion was one of confusion. IBM sold operating systems, so why would it champion a free alternative? If Linux didn’t become a success then IBM would have wasted lots of money – which could have been spent improving its other products. Alternatively, if Linux did become a success, pundits wondered whether that would result in a drop in IBM’s software and hardware revenues. Conspiracy theorists suggested that IBM would cause Linux development to split into hundreds of variants, weakening its ability to succeed in the marketplace. It all seems so fanciful now!

IBM established the Linux Technology Center in 1999 and announced Linux on the mainframe in 2000. Several Linux distributions (distros as old Linux hands refer to them) now run on a mainframe, including SuSE and Red Hat. Other distributions that run on a mainframe include Debian, Gentoo, Slackware, and CentOS. There are also thousands of ISV applications that have been recompiled to run on mainframe Linux. Importantly, Linux is not emulated on a mainframe, it runs as a completely native operating system – like z/OS etc. However, it is most often found running under the king of virtualization, z/VM, allowing multiple Linux systems to run on one lot of hardware.

There is a specialty processor available for Linux – like zIIp and zAAP – called the Integrated Facility for Linux or IFL. This processor runs Linux work, rather than running it on the normal General Purpose Processor (GPP), and so saves the user money because they are only charged for work running through the GPP. Of course, there is the cost of buying (in reality turning on) the IFL specialty processor.

It used to be that every sentence about a mainframe contained a mention of SOA, but nowadays the word is “cloud”. The IBM Web site says: “Cloud computing promises operational efficiency and cost reduction due to the standardization, automation, and virtualization. High flexibility, scalability, and easy management can be provided by the virtual Linux server environment on System z. Which basically says that Linux on System z is a good thing to have in your cloud computing environment because it is an efficient cloud computing platform.

So, happy 10th birthday mainframe Linux (or Linux on System z as we should be calling it). There are apparently 3150 Linux applications enabled for System z and IBM claims that 70 percent of the top 100 global mainframe customers run Linux – so I confidently predict that we will be celebrating its 20th birthday.

Arcati Mainframe Yearbook 2010 user survey

The Arcati Mainframe Yearbook 2010 has been available for download free from www.arcati.com/newyearbook10 for just over a week. Each new Yearbook is always greeted with enthusiasm by mainframers everywhere, and, not too surprisingly, there were over 2500 downloads on the first two days the 2010 edition was available. The Yearbook is always interesting, but especially interesting each year are the results of the user survey.

Respondents all completed a survey on the Arcati site between the 2 November 2009 and the 4 December 2009. 30% were from Europe and 46% from North America, with 24% from the rest of the world.

42% of the respondents worked in companies with upwards of 10,000 employees worldwide, while 19% had 1001 to 5000 staff. 16% of respondents had 0-200 staff and 12% had 5001-10,000. In terms of MIPS 45% of respondents had fewer than 1000 MIPS installed, 26% fell into the mid-sized category between 1000 and 10,000 MIPS, and 29% were at the high end.

Looking at MIPS growth produced some interesting results. Larger, more mature businesses (above 10,000 MIPS) were mostly experiencing some growth, but some growing by more than 50%. Sites in the 1000-10,000 MIPS range were typically experiencing growth of less than 10%. Sites below 1000 MIPS were also generally experiencing growth of less than 10%. The mainframe market does appear to be quite fragmented with competitive pressures at the lower end of the mainframe market.

With the environment and environmental issues getting so much coverage in the media these days, the survey asked whether IBM's recent green initiatives on things like power consumption and cooling had made the mainframe more or less attractive. Nearly three-quarters (72% - up from last year's 62%) said that IBM's green initiatives made no difference at all. One site said that the initiative made the mainframe less attractive. 19% felt it made the mainframe a little more attractive, and 7% felt it made the mainframe a lot more attractive. Clearly "greenness" isn't much of a selling point for mainframes.

With all the suing and counter-suing going on between IBM and NEON Enterprise Software over NEON's specialty processor software, the survey asked which specialty processors sites had. 37% of respondents had not yet activated or installed a specialty processor, 12% of sites had all three (IFL, zIIP, and zAAP), and a further 12% had two of the three specialty processors. More sites had zIIP processors (39%) than any other.

As usual, respondents were unhappy with costs. One commented: "The mainframe is too expensive (and perceived as being even more so). Costs for legacy applications must come down, and third-party software must support the mainframe." Another suggested: "The other platform costs may be increasing faster than mainframes but the starting point is drastically lower."

As always, there are wonderful nuggets of information in the survey.

The Yearbook can only be free to mainframers because of the support given by sponsors. This year's sponsors were CA, DataKinetics, OpenMainframe.org, Progress:DataDirect, and Software Diversified Services (SDS).