Sweating the details: distributed systems management
Advanced technologies bring to mind some conflicting truisms: "If
you can't explain it in a sentence, then you probably could do without it;" and, "For every complex question,
there is an answer that is simple, straightforward and wrong." These bring to mind distributed systems management
Have you ever you ever tried making the elevator pitch for managing decentralized servers and networks? The
first hurdle is that the idea may sound abstract to the CFO. Then try rallying the I/S faithful. Chances are, DSM
solutions mean different things to people who should know better.
DSM can only manage what is documented. The first question to ask is whether all relevant hardware, software
or network APIs assets have been identified.
This concept proved a major surprise for a regional health insurer, which operated a mix of financial and policy
applications, EDI and optical document management applications across two dozen Unix servers in several states.
It discovered the hard way that it had more system, application and network interfaces than it realized, and that
writing the software hooks tying those assets to its DSM framework was not the "plug-and-play" exercise
that it had anticipated.
Our ongoing study of over a dozen CA Unicenter and Tivoli Management Environment installations has yielded a
mixed bag of results: Successes mixed with efforts breaking down in midstream over a lack of commitment or preparation.
The obvious advantages -- reduced staff costs due to automation -- often do not tell half the story.
For the health insurer, the organization remained focused and the installation proved successful. While the
system reduced I/S administrator staffing by eliminating manual monitoring of servers, the system's improved alarming
and problem notification capabilities empowered operators to tackle wider responsibilities ranging from troubleshooting
to help desk support. Those improvements, in turn, hiked the I/S staff's salary expectations. The bottom line was
a mixed bag in labor costs. Still the company was more than happy to pay its operators premiums because their improved
knowledge added significant value to system and end-user operations.
Labor savings are also a deceptive factor for a mid-sized services organization, which used only the automated
report generation module of the DSM framework. The result was the amount of time necessary to generate and print
reports was cut by over 60%, saving roughly $12,500 annually in manual work previously performed by $25/hour operators.
There was a steep price to pay for those savings. Every Thursday report processing consumed so much server and
network overhead the Unix administrator typically spent the entire day stamping out brush fires. Those eight additional
hours of Unix administrative costs were more costly than those of the operators whose labors were saved.
The silver lining was the report generation tool finally proved the need for vast infrastructure upgrade. DSM
installations often point to the need for new investments that are not always anticipated. A midwest financial
institution installed job scheduling and event management, realizing modest downtime savings that failed to offset
the costs of licensing or implementation. Had it added related modules, such as performance management or automated
problem resolution, return on investment (ROI) might have turned positive. The culprit was not a lack of resolve
for buying software, but the inability to muster the justification for dedicating a specialist to focus on DSM.
But the rewards are there for the determined. A national wireless communications firm figures 80% reduction
in trouble tickets from its customer accounts application. This has not only reduced internal labor costs, but
helped attract new sales by easing the process of activating new customers. Now that's cold cash.
Tony Baer is principal with onStrategies, a New York-based consulting firm, and editor of Computer Finance, a monthly journal on IT economics. He can be reached via
e-mail at email@example.com.