Keeping your systems bang on schedule?

There were some things that ‘mature’ IT managers, those from the mainframe era, took for granted. One was job scheduling – not of manufacturing production, but of batch compute tasks. Scheduling software was used to automatically manage complexity – the timing of batch runs, key business and manufacturing applications, data queries, whatever – dynamically balancing demand against processor resources to ensure efficient operations. It was background stuff, but it was there because it was necessary and it worked. Enter client-server based on networks of non-deterministically coupled PCs (or similar) and servers, where computing is an expandable commodity. And the assumption has become that ‘the network’ will just deliver by throwing compute power at it. Today’s response to job scheduling software: “You don’t need that,” or “What is it?”. Indeed, best estimates are that today less than 10% of businesses have it. Most of the time that’s fine, because your IT infrastructure does keep doing what’s asked of it, and even if it doesn’t quite, shortfalls are pretty transparent to users. Only infrequently does something go wrong enough to be noticed and for questions to be asked about the network’s ‘availability’, or ‘robustness’. Control and repair However, today’s job schedulers can provide an excellent solution that can pay for itself very quickly – and IT professionals are beginning to think it should now be moving back up the business IT agenda. Because today, its value is not only in automatically running and overseeing jobs, but just as important, in restoring them. Its protection against failure, ensuring successful operations through fallbacks, alerts and workflow, is proving a big money and downtime saver. “Job scheduling is about ensuring business IT availability and robustness. Not having it can cost tens of thousands of pounds,” says Dave Thompson, UK managing director of French job scheduling software developer Orsyp – an HP and Sun supplier, with Peugeot-Citroen and Festo among its bigger clients. How? Being able to define paths to recover from errors means that downtime – resulting, for example, from a server going down overnight and causing a job to fall over so that pick lists, work-to lists, material allocations, JIT requirements, are not available for the next day’s operations – can be avoided. Sounds familiar? But handling the complexity and chaotic nature of open systems is another benefit. Thompson: “Unscheduled jobs, like web ordering, and batches hitting systems in an uncontrolled fashion – there’s an expectation of ever more [computer] flexibility. But if you hit systems when they can least afford it, there will be an impact.” And while many of the major ERP suites have basic job scheduling, users agree it’s rarely man enough. Jock Hunter, managing director of SBB, which has installed UC4 job scheduling software at, among others, Bosch, Siemens, Air Products and Pfizer, agrees. “Our systems take over where it’s simply too complex for human administrators.” The alternative is to run manually or have the IT department spend time building and maintaining scripts to do it. Cadbury Schweppes, which has finally gone live on SAP in Australia, is an SBB user. Cadbury has been working to consolidate 54 data centres and 27 instances of SAP to just three, with the stated objective of reducing to one. Craig Holmes, project manager, says, with the complexity resulting from thousands of SAP users, that couldn’t be achieved without UC4’s job scheduling and batch management. Similarly, Orsyp user Delphi’s IT manager Elsa Exbosito in Spain says it’s central to her operations based on Cincom ERP. “We run it every day. We have NT servers and VAX VMS servers and lots of applications for controlling the stocks, the manufacturing, that run every night.” And Pat Leach, CIO for drug discovery company Inpharmatica, concurs: “It’s absolutely critical to what we do.” The alternative, he notes, is redundancy, which is going to be inefficient and costly. So what does it cost? Thompson says that for a system involving say one NT machine you’d be looking at “around £5,000 all up.” For a manufacturing SME, the reality will be around £20,000–£50,000. “You could cost justify that on the basis of just one major problem a year,” he observes.