Mainframe columnist Robert Crawford addressed the high cost of third party mainframe software in a recent article. Here is a sampling of reader responses:
The dwindling cycle
Robert, I totally agree with you. This has been going on years. Cost is often sited as a prime reason for getting off the mainframe.
I believe their are two main drivers that tend to feed one another.
There are large mainframe shops out there but I don’t believe there are as many mainframe shops since the client-server movement. Corporate mergers and acquisitions have reduced the number of traditional mainframe shops too; in some cases maybe even more so than the so called demise of the mainframe.
Meanwhile we have seen much consolidation amongst the mainframe software vendors themselves.
So what we now face is fewer vendors (thus less competition) selling fewer copies (due to fewer shops) while attempting to grow their bottom line. Unfortunately it can tend to drive more customers to alternative platforms which sustain the downward spiral of fewer customers to cover the vendor’s costs of maintaining the software plus whatever greed factor may or may or not come into play.
Poorly designed mainframe applications the real culprit
This debate over mainframe costs seems irrelevant to the real world.
In the 30+ mainframe shops I’ve been in, the real cost is in poorly designed applications.
Modern busienss class hardware and systems software is intended to use relational
databases in a set processing relational way. Parallelism and other modern features
are expected to be exploited by the application design. zIIPs and zAAPs are now
expected to be exploited by the application design.
But, in reality, much processing uses the relational database in a hierarchical design
with the low cardinality keys at the top of the hierarchy controlling the processing.
Or even worse, the application design is a sequential flat file design. Read one record
from Table A; Take a column in that record and use it as a host variable to read a
record in Table B; and on and on down the chain of sequential processing.
The poor design of business applications is the single biggest cost in any large shop.
Discussions of hardware cost and vendor/systems software cost tied to peak CPU usage
is really quite irrelevant to reality.
A well designed application on an average platform will beat a poorly designed application
on the best platform.
What’s your take on mainframe software costs? Leave us your feedback in the comments section.