As an IBM mainframe user, what should be on your wish list for 2010? Most of my wish list items come from the things that aggravated me at work last year.
1. Fewer DB2 deadlocks and more concurrency
Along with the splendor and majesty of Parallel Sysplex came something called "sympathy sickness." Sympathy sickness occurs when an ailing LPAR manages to bring down its otherwise healthy brethren. IBM and customers have gone to great lengths to serialize and protect global resources. Now it seems like most Sysplex-wide slowdowns and lockups begin with databases. While I understand database locks will always be with us, I do wish the DB2 developers would find some better locking mechanisms.
Other impediments to full availability are database re-orgs and plan rebinds. Database utilities have come a long way since the days of weekend offline image copies, but we should be approaching the days when online transactions can continue to run no matter what's going on. This goes for system tables as well, when, for instance, an enterprise needs to rebind plans and can't afford to stop online traffics.
2. Better database diagnostics
Until IBM works out how to reduce DB2 timeouts, a better way to figure out how the things get started would help. IMS does a good job by recording conflicts in its logs. Finding DB2 conflicts, however, often requires long, painstaking and error -prone searches through job logs or accounting records. DB2 ought to record the lockups in its log or even another dataset for that specific purpose. Enterprises running complex applications that mix IMS and DB2 database calls would appreciate a utility that can merge DB2 and IMS lock information into a narrative explaining how the lockup started.
3. More record-level sharing documentation
Record-level sharing (RLS) is a black box with very few knobs to twist. The good news is it does its job very well with little fuss when things go well. When things start to go out of kilter, however, the lack of documentation and understanding leaves a lot of people scratching their heads. At that point, gun-shy system programmers may resort to gingerly poking at RLS by tweaking a few parameters and measuring the effect in test. The dread only gets worse when making more or less blind changes to production, where the stakes are much higher.
The documentation needs to be clearer, especially regarding when and where changes take effect. More discussion about cache (memory and coupling facility) and the locking structures would be appreciated, too.
4. More "legal" ways to use specialty engines
In the past I've expressed my frustration over getting workloads onto the specialty engines. Some of the rules seem specifically designed to keep work off of them. Then there's the so-called "generosity factor" governing the maximum amount of specialty engine a workload can use no matter how eligible it might be.
Originally, IBM created the specialty engines to coax distributed workloads that were otherwise ill suited for big iron onto the mainframe. However, I think this would be a good time for liberalizing the rules to include more traditional mainframe workloads. IBM would reap a lot of good will from such a move. I'm also sure most customers would rather boost their specialty engine use by following the rules rather than antagonizing IBM or buying some sort of magic software.
5. A change of attitude at IBM
Up to now IBM has fought threats to its mainframe empire through case studies, lawyers and buyouts. Perhaps the biggest threat was last year's introduction of zPrime, which allows heretofore ineligible workloads to run on specialty engines. IBM responded by bullying customers.
If the customer base of the z/OS platform (as opposed to just mainframe hardware running zLinux) isn't eroding, it certainly soon will be. To shore up z/OS, IBM needs to undergo a change of attitude. I recommend reducing prices so the platform's value is self evident without sales pitches full of complicated hand waving about economies of scale, power consumption or ease of maintenance. Instead of bullying customers, work with them to reduce costs in real dollars to make the platform attractive within an enterprise. Stop suing competitors and begin working with them to expand the mainframe's world instead of shrinking it. Finally, I give IBM the best piece of advice my mother ever gave me: "Be nice!"
ABOUT THE AUTHOR: For 24 years, Robert Crawford has worked off and on as a CICS systems programmer. He is experienced in debugging and tuning applications and has written in COBOL, Assembler and C++ using VSAM, DLI and DB2.
What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at email@example.com.