The Future of The Mainframe: Gary Barnett
The Future of The Mainframe: Gary Barnett
Gary Barnett
October 2005
1
THE FUTURE OF THE MAINFRAME
Key messages
The mainframe is alive and well as a platform for very
large workloads
The mainframe is alive and well at the high end. IBM is keeping the
promises that it made in its mainframe charter and continues to deliver
steady improvements in price/performance. At the very high end (in the
‘more than 2,000 MIPS’ category) IBM’s new mainframe, the z9, has done
a lot to secure the future of the zSeries.
independent software vendors and their desire to lower the total cost of
acquiring their solution. In other cases, migration will be driven by the
selection of a brand new application that runs on a lower-cost hardware
platform.
There is no ‘crisis’ (although with ongoing budget pressure it may feel like
that at times). The mainframe is not going to disappear any time soon, nor
is IBM’s commitment to it.
The ‘skills crisis’ that some commentators claim will bring the world of
Cobol to a halt is at least a decade away, so there is no need for urgent
action. However, employers need to consider the changing needs of a
developer population that is undoubtedly getting older.
You have choices, and the smart suppliers will be the ones that can help
you take different approaches. Beware of the supplier that only talks about
migration as opposed to legacy renewal.
would allow you to continue selling for the mainframe (using the
mainframe’s Linux support), as well as the wide range of hardware
platforms on which Linux runs.
Currently, the choice that many see is, ‘stay with IBM on the mainframe,
or move to non-IBM hardware’. Plainly this doesn’t make sense – IBM has
a wide range of non- mainframe hardware offerings in the form of pSeries,
iSeries and xSeries. Surely if people are going to move off the mainframe
platform – and a growing number will – it makes sense for IBM to help
them move to one of its non-mainframe platforms rather than to a
competitor’s?
Once again, the message here is that if IBM doesn’t, then someone else
will.
Many legacy applications and systems are IT assets that have a particular
combination of attributes: they manage resources that are crucial to the
operation of the business, and they are seen as inflexible or brittle in the
face of technological and business change.
Increasing numbers of large businesses (and these are the ones which
tend to host legacy systems) are looking to make significant adjustments
to their IT portfolios in order to align IT expenditure and operations with
business priorities. It is clear from this that if these grand IT
transformation projects fail to properly address legacy applications and
systems, they themselves may well fail to deliver value.
Analysis
In general terms, the mainframe is fit and well. Brand-new mainframe
MIPS shipments (as opposed to replacement MIPS) continue to grow by
around 15% a year, and the number of lines of Cobol code in production
continues to grow by between 3% and 5%. The mainframe isn’t going to
disappear tomorrow; indeed, we expect mainframes to be in production
well into the next decade at least.
Over the next five years smaller mainframes will come under intense
pressure from x86-based (Intel and AMD) platforms as their
price/performance improvements out-pace the mainframes by a significant
margin. The price/performance gap is still widening, despite the excellent
progress that IBM has made with the mainframe.
IBM is the leading mainframe vendor, with over 70% market share.
Therefore, while this report covers the global mainframe market – it is
largely dominated by IBM.
• IBM is committed to its mainframe technology: the recent release
of the z9 (IBM’s latest generation of the zSeries platform)
demonstrated the fact that IBM is committed to supporting its existing
clients and sees a long-term future in the platform. IBM has invested
well over $1 billion in developing the z9, and has committed to
continue investing heavily in mainframe R&D.
• IBM is improving the economics of the platform: IBM’s mainframe
charter represents a commitment on IBM’s part to its mainframe using
clients. To drive down the total cost of the mainframe platform (that is,
the cost of the hardware, software and administration) and to make it
a first-class citizen in the (comparatively) new world of distributed
computing.
Price
Cost of integration
Cobol skills
There is growing concern among our client base about the availability of
Cobol skills. We do not believe that there will be an acute shortage of skills
in the short term, because the fall in the number of Cobol developers will
be offset by a comparable reduction in the amount of Cobol development
that takes place. Over the longer term, we do see a shortfall in Cobol skills
in the 2015–2020 timeframe as the older generation of Cobol developers
(in the 45–55 age range) leaves the workforce.
Cost
Innovation Risk
Source: Ovum
We also use this approach when talking to vendors about their offerings –
if a salesperson cannot explain how the product will help to address at
least one, but preferably all three, of these issues, then it’s going to be a
tough sell.
Cost will never cease to be a key priority for IT managers. While the
painful cost reduction targets of 2001/2002 may be behind us, every CIO
that we speak to talks about continued downward pressure on core IT
spend. Many are now reporting that there is new money flowing into IT –
but this new investment is, more often than not, tied to very specific
business programmes and goals.
With cost still taking a high priority, the question then becomes, ‘how do I
lower costs to free-up resources to support innovation?’ Essentially, the
task is to fund the innovation budget by saving costs elsewhere.
The key to this lies in understanding your current cost base, and the
contribution that your key expenses make to the business.
We don’t pretend that this is ‘rocket science’ or that it’s a particularly new
idea, but we do use it as a framework when looking at how our clients
need to develop and validate strategy.
Delivery Utility
Spend
Assembly Assembly
Utility Delivery
Source: Ovum
Although the balance varies according to the individual client and the
sector that the client operates in, around 60% of IT budget is spent on
delivering and maintaining utility activities, with the assembly and delivery
categories left to fight over the remaining 40%. Once we look at spending
patterns, we notice that the greatest proportion of IT spend is frequently
directed at those activities that differentiate the company the least. Part of
the process of legacy renewal and modernisation is the modification of this
balance – so that, ideally, the bulk of IT spend is directed at differentiating
activities.
Of course, many of them have good things to say about the mainframe as
well – the mainframe is still regarded as the most reliable and secure
computing platform there is. However, increasingly, the cost and lack of
Interestingly, very few of our CIO clients believe that the mainframe is
going to ‘die’ or ‘disappear’ at any time in the foreseeable future, but the
key challenges remain.
• How do you lower the cost of running legacy applications?
• How do you make it easier and less expensive to modify and enhance
legacy applications?
• How do you make it easier and less expensive to integrate legacy
applications with the rest of the IT system?
• integration
• re-engineering
• platform migration.
Source: Ovum
The vast majority of this decline will come from small mainframes (sub-
100 MIPS machines). Indeed, we expect the number of larger mainframes
(those supporting workloads of 1,000 MIPS and over) to grow.
We’ve divided mainframes into different classes in terms of their size (as
measured in MIPS):
• 0–100 MIPS – tiny
• 101–500 MIPS – small
• 501–1,000 MIPS – medium
• 1,001–4,000 MIPS – larger
• 4,001+ MIPS – super-sized.
By contrast, we believe that the number of larger mainframes (in the large
and super-sized classes) will grow slowly over the same period. By 2010,
we believe the distribution of mainframes according to our classification
will be:
• more than 4,000 MIPS – 5%
• 1,001–4,000 MIPS – 13%
• 501–1,000 MIPS – 31%
• 101–500 MIPS – 28%
• less than 100 MIPS – 23%.
Source: Ovum
Mainframe migration
The mainframe migration market will grow gradually over the forecast
period, peaking in 2009 before beginning to fall off.
Number of replacements/rewrites
2005 2006 2007 2008 2009 2010
> 1000 2 3 4 4 5 5
501 - 1000 mips 4 6 7 8 9 9
101 - 500 mips 89 97 86 89 90 81
< 100 mips 216 239 269 283 280 264
Total 310 346 366 383 385 359
Source: Ovum