subscribe: Daily Newsletter


The future of the mainframe


João de Oliveira, sales director at MigrationWare, weighs up the pros and
cons of the mainframe

Technology writer Stewart Alsop predicted in 1991 that the last mainframe
would be unplugged in 1996. The prediction did not come true and the
mainframe powered into the 21 century. In 2004, the IBM mainframe celebrated
its 40th birthday and IBM unveiled the z9 mainframe which was the result of
over $1-billion of investment in research & development.
The mainframe computer has therefore retained its relevance for over four
decades, and over the last two years IBM and software vendors like Computer
Associates (CA) have even gone on drives to promote mainframe computing.
There's no faulting the mainframe when it comes to reliability, security and
the processing of large scale, high volume transactions. What the mainframe
does, it does very well. And according to Gartner, mainframe growth is not
slowing down in large corporates either.
The research house estimates that large mainframe users have been increasing
their mainframe environments – measured in millions of instructions per
second (MIPS) – steadily over the past few years. Most of these corporates
will increase their installed MIPS at a compound annual growth rate of
between 15% and 20% over the next two years.
CA has stated that corporates with large mainframes are increasing their
capacity by between 20 percent and 50 percent per annum.
This is also seen in research released by Ovum which states that companies
with over 1000 MIPS (large installations) will be increasing their installed
MIPS over the next two years, while medium sized installations with less
than 1000 MIPS will be decreasing the amount of MIPS.
However, there are challenges that the mainframe faces. The Ovum research is
certainly an indication of the larger mainframe users increasing their
mainframe usage, while smaller mainframe sites are decreasing their
mainframe usage or migrating to other platforms.
Mainframes are also expensive, retailing at just under R1-million, and
upgrades on large mainframes can literally run into the tens of millions of
rands, as can the software licensing costs, which are rated by Gartner as
the single biggest inhibitor to mainframe growth.
Development is done on the mainframe, which means there is often a conflict
between transactional priority and development priority. Software
development is often sidelined with mainframe business transactions taking
There is also a shortage of developer skills. Locally, there are no longer
any Cobol training institutions and the banks and other financial
institutions are training their own Cobol developers as older Cobol
developers retire.
And while MIPS stats are on the rise, Ovum has stated that the number of
mainframes in operation would decline by a third in the five-year period
prior to 2010.
Given the previously stated advantages of the mainframe, there is little
chance that the large corporates will give up using their mainframes anytime
The mainframe's future therefore lies in streamlining mainframe usage. The
first way to do this is to get the development off the mainframe. Companies
can shift the development environment into a more modern environment to free
up capacity on the mainframe, giving developers more advanced toolsets to
work with.
Expensive mainframe upgrades can thereby be delayed or averted as the
mainframe is used only as a final systems test and deployment server.
Additional systems, like information centres which store transactional
history, can also be pulled off the mainframe. This also frees up capacity
on the mainframe and lowers the cost of ownership.
Even the large companies are looking at the option of moving certain
applications off the mainframe into a Unix, Linux or Windows environment. A
"lift and shift" approach is used in this case where Cobol applications are
adapted with minimal or no modifications to be able to access data from a
new database on the new platform. The process is called modernisation.
Mainframe behaviour is also emulated in the new environment to ensure
applications are able to work effectively. In many cases performance is
This reduces the average elapse time required to process the same amount of
data on the mainframe and allows companies to utilise business intelligence
software to analyse the data in real time, as opposed to the delayed or
batch reporting processes offered by the mainframe.
Applications written some time ago therefore don't necessarily need to be
replaced through expensive rewrites or packages. Modernisation moves
enterprise applications into the 21st century through the use of modern
programming tools.
Some companies with smaller mainframes are moving off the mainframe
environment onto mid-range or Wintel platforms, often through the use of
packages or "lift & shift" migrations. Applications written in aging 4GLs or
other proprietary environments can be converted to the more widespread and
platform independent Cobol language using automated processes. This trend is
unlikely to reverse and will in fact accelerate as more and more smaller
mainframes are left with one or two applications bearing the entire
mainframe cost.
It is also unlikely that smaller organisations will migrate onto the
mainframe given the costs involved and the fact that there is a real
perception in the market that mainframes are only being used by companies
which have had them for 30 years.