The Reality Facing The Mainframe World

An Acxiom White Paper
The Reality Facing the Mainframe World
by George Drabik, Acxiom
Dennis Ford, Acxiom
Jeffrey Shoup, Acxiom

Acxiom | White Paper

You manage a mainframe environment that runs one
or more of your business’ mission critical applications.
Things are good; security, performance and reliability are
just where they should be. But when you think about
your long term staffing strategy, you cringe, because
most of the people on your mainframe staff have long
term plans that include traveling, fishing and gardening
… in other words, retirement.
Steward Alsop wrote the following in a March 1991 edition of InfoWorld, “I predict that the last
mainframe will be unplugged on March 15, 1996.” But the fact is that the mainframe’s future
today is about as bright as it was in the 1970’s. Emerging enterprise-wide architectures such as
Business Integration, Virtualization, SOA, along with zIIP, zAAP and IFL engines and hipersockets
are creating greater potential value in the mainframe than ever before. And in the March/April
2008 issue of Mainframe Executive, publisher Bob Thomas wrote, “Consider this, a single z10
system can do the work of up to 1,000 Intel servers, reduce energy costs by 80 percent, reduce
floor space by 85 percent, and conserve labor costs by 80 percent.” The facts are this:
Mainframe MIPS are growing; new technology continues to be introduced into the platform;
z/Linux is doing well and continues to gain clients; and we see a continued improvement in the
cost per MIPS to run the platform. The mainframe is rebounding dramatically.

But the news is not all good. According to Share board member Robert Stanley, “For every two
mainframe workers reaching retirement age and leaving the workforce, only one new mainframe
worker is being added.” A February/March 2005 z/Journal article indicates that “a recent META
Group study found that 55 percent of IT workers with mainframe experience are over 50 years
old.” These same workers hold so much information and have so many skills that replacing
them would be difficult even if the mainframe was perceived as a technology with a future. Sadly
however, there seems to be a perception by college students that the mainframe platform is not
a viable long-term career choice. And yet, mission critical applications remain on the mainframe,
and those trying to move to another platform are finding the process challenging, expensive,
and time consuming.

1


The “last app standing” dilemma
Companies that try to move off their legacy mainframes are confronted with the “The Last App
Standing” dilemma. It’s that one application whose functionality isn’t part of the new client server
software suite you purchased, the one that has your custom business logic in it, the one that’s
written in COBOL and VSAM that would take man years to port to Java and SQL, but it’s also
the one that provides a key part of your business functionality. “The Last App Standing” is one
good example of when it’s a good time to consider alternatives to your current IT paradigm. It
would certainly be too expensive to maintain enough critical mass to continue to operate your
own mainframe environment, but you still don’t have a good replacement for that function.

Maintaining critical mass
There is another reality facing companies that are migrating away from the mainframe. Those
that choose to steer toward client server technology — save some long established and
business critical applications — confront the issue of “critical mass.”

In physics, critical mass is the minimum amount of nuclear “fuel” needed to sustain an ongoing
reaction. If the “fuel” drops below critical mass, the reaction shuts down, the reactor goes cold
and the lights go out. In an IT shop, critical mass is the minimum level of resources — software,
people, hardware and facilities — that are needed to maintain a production data center.

For shops that are planning to migrate their processing from the mainframe onto client server
platforms, critical mass is manifested as a floor to the cost structure; if your MIPS are decreased
by 50 percent, oftentimes it turns out your mainframe costs hardly decrease at all. For many
shops whose direction is to move to new technology platforms, the problem becomes
determining how to maintain your critical mass as you try to downsize your old environment
while migrating to the new. The problem is that some costs simply don’t decrease in a linear
fashion, or as usage decreases.

“Right sizing” means much more than headcount
Managed mainframe vendors have traditionally priced mainframe CPUs on a “pay for the MIPS
you get” model. This goes back to the days when building a faster mainframe involved using
higher cost components, and it required a larger chassis using a 32-bit backplane instead of a
16-bit or 8-bit backplane. The reality of today’s mainframe is that the only difference between a
25 MIPS Z10-BC processor and a 1,500 MIPS Z10-BC processor is the capacity marker that
IBM supplies on the microcode diskette. On the current generation of mainframes, the CPU

2


board comes fully populated with processor engines, but between one and five of the engines
will be active for the z/OS workload. So, the basic frame with its card cages, power supplies,
backplane and CPU board sets a floor on the manufacturing costs for a mainframe that is no
longer driven by the number of MIPS a machine is designed to deliver. Economically, this means
that the fewer MIPS required on any given processor, the higher the per-MIPS cost becomes.
The net result for companies trying to downsize their mainframe environment is that they can
spend a lot of time moving applications off the mainframe to reduce their MIPS by half, only to
find that the scaled down CPU can cost nearly as much as the CPU they’ve replaced.

Other things that don’t scale down
How do you run half of a shift when your workload shrinks? How do you effectively retool your
aging mainframe support staff when they are needed less than full time but still need to keep
their skills current? The toughest things to scale down are the resources and support necessary
to “keep the lights on.” While this is not impossible, it can be tough.

Typical data center workload requirements represent a continuous spectrum of staffing needs,
but the actual staff available is quantized in units of one person. For example, suppose you’re
running your online interactive workload during the normal business day, and you’ve scaled
back your applications so only four hours worth of daily batch work remains for the second
shift — backups and re-orgs that must be completed each evening. For operations, let’s say
that this requires a staff of 1.5 people to complete, but where will you find half-a-person?
Typically, skilled IT staff aren’t going to accept part time positions, so you will need to retain an
operator for both shifts and then try to figure out what other projects and tasks can be used to
fill in their time. Hopefully, as your transition to a new platform is ramping up, the mainframe
workload ramps down, but it’s more likely that you will have to fully staff the new environment
before you are able to cut back on the old.

Staffing for seven-days-a-week operations also presents a problem — if you have enough work
you can try split shifts of three-day / four-day weeks with 12 hour work days to cover all seven
days, but if there isn’t 12 hours of work per shift, you’re back to trying to find part time
operators to cover the weekends. An 8x5 operation is one plateau, and 24x7 is another, but
anything in between is difficult to scale. In particular, if you’re trying to downsize your current
24x7 operations, you may find that as you reduce workload, your operations cost does not
scale down until you can eliminate an entire shift. Your cost to run 500 MIPS of workload 24x7
may not be much different from your former 1,000 MIPS of 24x7 because you have to maintain
the critical mass to run 24x7 regardless of the number of MIPS you’re supporting.

3


A brighter future?
Given the challenges facing the mainframe — a retiring workforce, migrating mission critical
applications, the economics associated with migrating off of or reducing MIPS — what is the
future for the mainframe? Credit Suisse market research states, “We believe that the mainframe
market has reached a point at which ‘the law of small numbers’ comes into effect … growth of
MIPS from the customers remaining on the platform more than offsets the loss of MIPS from
those migrating off. …”

Here are some estimates related to how embedded the mainframe is in business today
according to some very reliable sources:

• “200 billion lines of COBOL code in existence” – eWeek
• “5 billion lines of COBOL code added yearly” – Bill Ulrich, TSG Inc.
• “Between 850,000 and 1.3 million COBOL developers” – IDC
• “Majority of customer data still on mainframes” – Computerworld
• “Replacement costs $20 trillion” – eWeek

To meet the needs of the future there is an effort to replenish mainframe workers with new
blood. A Wired Campus report from April 7, 2008, notes that IBM “has furnished Illinois State
University with its System z890 in order to train more students in mainframe technology.” This
and other such efforts to replenish the mainframe workforce are being monitored closely by
those who envision a bright future for the mainframe.

For an average mainframe data center (excluding the network), roughly 40 percent of your costs
are for software, 30 percent for salaries, 25 percent for hardware and five percent for facilities.
Some strategies you can look at to keep that mission critical operation running for the long term
include: working out a plan to keep retiring workers longer, offering part-time opportunities;
beginning knowledge transfer sooner (now), well before it’s too late; making your own
connections with local colleges to get interns and graduates; making that challenging decision
to move to a platform that appears to have more readily available support; or taking advantage
of various outsourcing options.

4


How one company is solving the problem for clients
Acxiom, an IT Services provider with competencies in global interactive marketing, was founded
on mainframe processing. Nearly 40 years later, the company continues to recruit and retain the
most highly-skilled, experienced mainframe professionals in the nation to address clients’ most
complex mainframe challenges. Additionally, the company has consistently delivered, on average,
20 percent cost savings for clients — a must in today’s economically troubled environment.

How can Acxiom continue to deliver mainframe MIPS at a lower cost and with less risk? It boils
down to scale and leverage across many customers and workloads, avoiding the problem of
critical mass for those customers who wish to decrease the size of their MIPS. In addition,
Acxiom has developed programs to recruit mainframe skills right out of college and has even
partnered up with a software vendor developing a shared “Mainframe college.”

Another approach being taken by Acxiom clients experiencing the retirement purge is to have
their environment remotely managed. Remote Infrastructure Management (RIM), mainframe or
other, offers clients technical support at whatever level they choose, and operational support for
functions within their environments that do not require touching the hardware. CIO Magazine
made the following 2008 prediction: “To date, this segment (RIM) of the outsourcing market has
been growing at about 20 percent a year, according to neoIT. The Everest Research Institute
predicts the RIM market will pick up to the tune of 60 to 70 percent growth in 2008. Besides
increasing comfort levels with the RIM concept, neoIT also suspects that an economic downturn
in the U.S. could fuel greater demand for the lower-cost infrastructure outsourcing option. …”

The conclusion? The mainframe is alive and processing. As a mainframe executive, various
solutions are available to help address your challenges and meet the needs of your business.
And to help you sleep better at night!

5

See how Acxiom can work for you.
For more information, visit our website at
www.acxiom.com/itservices or call:

1.888.3ACXIOM

AC-2011-08 ELEC 12/08

The Reality Facing The Mainframe World

More Related Content

What's hot (20)

Viewers also liked (8)

Similar to The Reality Facing The Mainframe World (20)

The Reality Facing The Mainframe World