Consider for a moment just how insignificant—or surprisingly sizable—1 percent can be:
• 1 percent of your daily calorie intake
• A 1 percent raise
• 1 percent of the national deficit.
But what about 1 percent of an organization’s CICS transactions? And what if that 1 percent is underperforming?
On the surface, 1 percent is very good when measured in terms of MIPS consumption. Interpreted this way, 1 percent really means Service Level Agreements (SLAs) are being met—a pretty good day’s work in the mind of a mainframer.
But let’s explore the 1 percent in a different way.
Take a large financial institution that had long-running DB2 SQL statements. So long, in fact, that it would be really, really difficult to purposely make a DB2 call run this long. And this was happening in production. Why wasn’t it caught earlier?
Here’s why. When the load tests were run, the CICS administrator used the CICS monitor to watch the CICS portion of the test. The CICS transaction was invoking the DB2 SQL, so, if the DB2 SQL ran long, the CICS transaction ran even longer. Why hadn’t the CICS administrator captured this? The answer was that 99 percent of the CICS transactions had a great, sub-second response time. One percent of the transactions did not. The CICS specialist simply thought that 1 percent was an acceptable level of performance.
Seems logical, right?
But what if the organization runs approximately 1 million transactions each day and 1 percent of those transactions took multiple seconds to complete? That’s 10,000 transactions. Now do you think that’s an acceptable threshold?
Framed another way, what if a customer in the grocery checkout discovered she didn’t have enough money in her account to complete the transaction. When she used her iPhone to transfer funds, the poor-performing CICS transaction caused her to wait minutes for the money to transfer, embarrassing her and annoying customers behind her. Is that acceptable? Not to the customer.
In today’s highly competitive marketplace, where companies—particularly banks—vie for the same customers, allowing 10,000 underperforming CICS transactions, that is, 10,000 customers to be negatively impacted every day, is a huge risk.
Still, in spite of the risk, the CICS administrator was quite comfortable with the 1 percent. Why? Call it a kind of myopia, perpetuated, at least in part, by generations of mainframe professionals. But don’t blame them.
Decades ago, customer relationships weren’t at stake if 1 percent of mainframe transactions were underperforming. A bank’s mainframe was mainly accessed by tellers, loan officers and other internal personnel. End user performance wasn’t a concern because the mainframe didn’t impact the paying customer. The focus—and rightly so—was on systems monitoring and managing costs.
Today’s mainframe professionals are doing much the same despite the significantly different role of the mainframe in enterprise IT. CICS administrators are concerned with monitoring the CICS region. DBAs are focused on DB2 and z/OS professionals … . You get the picture. But this myopia is a detriment to today’s enterprise. It’s no longer “business as usual.” Mainframe professionals should embrace the dual nature of mainframe performance: CPU performance (cost efficiencies) and the end-user experience (revenue opportunities) to maximize the value of the mainframe.