At age 28, Boas Betzler is known at IBM as the "grandfather" of Linux on the mainframe. In mid-1998, he began a port of Linux to the IBM zSeries, and he's shepherded the Linux mainframe project from the beginning to the first shipments.
In porting Linux to an IBM mainframe, Betzler went against the grain both internally at IBM, where many saw Linux as a competitor to IBM's own mainframe operating system, and among many analysts outside of IBM, who suggested the port wasn't possible without a huge investment.
Betzler's "fun" project has helped IBM record double-digit revenue growth for the first time in more than a decade, "driven in significant part by Linux on the mainframe and its ability to consolidate literally hundreds of Sun and HP
servers for customers seeking savings on energy, floor space and management
costs," according to IBM's PR team. In the fourth quarter of 2001, 11 percent of the mainframe computing capacity shipped by IBM was slated for Linux applications.
Linux on the mainframe remains controversial. In late February, Shahin Khan, chief competitive officer at Sun Microsystems wrote a position paper, Linux on the Mainframe -- Not a Good Idea. In the paper, Khan basically argued Linux was the wrong tool for the job on a mainframe.
Khan's paper prompted a response from SWsoft chief scientist Alexander Tormasov and a response to the response by Khan earlier this month.
Betzler, now a member of IBM's Linux Technology Center, is the lead strategist for
embedded Linux. We asked Betzler, part of IBM's senior technical staff, about that controversy, about his original idea of porting Linux to the mainframe, and about where he sees Linux going in the future. [One note: His answers were run through IBM PR before getting to us.]
Question: What made you think that Linux would be a good idea on a mainframe?
Betzler: I knew the mainframe was years ahead of other servers in terms of
reliability and scalability. It became clear that this world-class server
needed a world-class operating system to run the new applications that make
up the Internet. Linux was the obvious choice since it had already proven
to be portable and ubiquitous. Additionally, university students were
becoming increasingly well versed in Linux. When you grow up with personal
computers and UNIX systems, you just expect to get the same look and feel on
the big servers. That is what Linux offers for the mainframe.
The partitioning and virtualization of the mainframe opens up a whole new
world. With these capabilities, several hundred operating system images can
run parallel to one another on the same box. IBM's experience in the server
field has allowed us to consolidate a soccer field full of servers into
logical images on a single box.
Question: What's the advantage of Linux on a mainframe over other Oses?
Betzler: Linux is a straight-forward implementation of a UNIX-like operating
system. Linux developers optimize for performance. The traditional mainframe
operating systems are much richer from a functional standpoint, include more
service features and provide cleaner isolation between the processes in a
single operating system image. With Linux, mainframe users now have a choice
to enable people with UNIX skills to deploy Internet applications or
consolidate servers. If they want to run a mission critical transaction
system or build a large, clustered, single image database, however, they
will probably chose z/OS.
Question: How long did it take to get Linux working on a mainframe? How many
people were on your team?
Betzler: When we started the project, we already had a working version of the GNU compiler that produced code for the mainframe architecture. It took us about
six months to port most of the kernel to the stage where we could see the first
kernel messages on a local console. About two months later, we had the first
user processes running. The first port was done by five excellent coders who
did most of the work in their free time. They were amazing. One Monday,
one of our coders told me that he had ported the complete GNU assembler over
to the mainframe architecture over the weekend, a job we had previously
dismissed because I thought it would be a major effort. We had a proprietary
assembler before that, but as he said, "Why shouldn't I do the work if it
only takes me a day?"
Question: How much resistance did you get to the idea from internal IBM people?
How much convincing was needed to get the go-ahead with the project, and
then to actually market Linux on mainframes?
Betzler: In the beginning it was clearly a skunk work project and operated under the cover of senior technical people who gave us the opportunity to prove
the feasibility of the project. Many experienced people told us that the
project was doomed for technical reasons, or [who] maintained it wouldn't be
attractive to customers. It required a year of strong conviction and
creative approaches to drive the project through the company. When the
project was shared with the community, however, we experienced hundreds of
downloads within the first few days. IBM is a customer-driven company, so
when we recognized the interest, we moved forward. We are not a company to
sit back and protect our old business model. We are focused on the
marketplace and go where our customers go.
Question: What was your response to the recent statements coming out of Sun that
say Linux on a mainframe doesn't make sense?
Betzler: They say they have a business case that Linux on mainframes doesn't
work. Well, this is a surprise, because IBM and many other independent
consultants and consulting groups have business cases that show it does.
The key difference between Sun and IBM is not Linux on the mainframe, it is
IBM's commitment to Linux. From Intel thin servers, to department servers,
to RISC-based enterprise and mainframe class servers, IBM has committed and
delivered Linux support across its entire range of hardware.
Question: What are you working on now? Working on embedded Linux must feel a lot different than porting Linux to mainframes.
Betzler: I feel changing your point of view from time to time is important. Now I am focused on making Linux a better fit for constrained devices and purpose
optimized systems. It's true, cell phones, wrist watches and game stations
are very different from mainframes. But the beauty of Linux is that I can
deploy the skills I have already acquired. And of course the reuse of
components and functions allows you to avoid re-implementing the same
function over and over again.
Question: Your bio says you're "one of the key players who turned the whole company into a new direction." How's it feel to have helped IBM make this big change
to embrace Linux?
Betzler: I always knew that Linux on the mainframe would make sense. But the
impact that my work had still amazes me. It also demonstrates how much this
company has changed and how its technical community offers young people
exciting opportunities. Above all, it is pretty cool and a lot of fun.
Question: Do you see other places where Linux isn't widely used that could be new
markets for Linux?
Betzler: We all know that small devices will become more powerful and useful over the next couple of years. Information devices like PDAs, cell phones and
alternative clients all need an operating system as infrastructure. Linux
can be used to power these devices and is already used in PDAs and storage
Question: What do you see as the next big push for Linux?
Betzler: Short term, I see many servers being used as appliances and these
appliances turned into computing utilities. The focus on these systems is
not the system software but the application, the service it provides. The
more developers have to focus on real functional value in the application,
the more they will pick a commodity operating systems that follows open
standards. Linux is that operating system. The next step will come when
Linux is deployed on consumer devices like digital video recorders or game
stations or even part of appliances like soda machines.