Some people might smirk at the thought of calling job schedulers some of today's hottest systems tools. But the new breed of multiplatform, XML-enabled, event-driven job scheduling tools represents several leaps forward from the old-fashioned IT workhorses of the mainframe era.
Take a look at some of the organizations using them:
Narex Inc. in Golden, Colo., is using Tidal Software Inc.'s SysAdmiral job scheduler to automate and halve the time it takes to process credit reports.
The Wisconsin Department of Corrections in Madison used Argent Software Inc.'s Argent Job Scheduler to run jobs just 20 minutes after installing the software.
In Dallas, BMC Software Inc.'s Control-M job scheduler has let the city's data center trim a round-the-clock staff of 48 to a mostly daytime staff of 28.
"After decades of being viewed as drab mainframe tools, job schedulers have become sexy," says Patrick Dryden, an analyst at Meta Group Inc.'s Houston office.
Job schedulers were designed as simple job launchers first for mainframes, then for Unix systems. A master command would direct multiple agents to run reports, update databases and perform similar tasks at set times.
Over the years, enterprise programmers have written job scheduling policies with interdependencies. One policy might state that Job 2 (running reports) shouldn't start until Job 1 (updating a database) has run successfully. And that was about as complicated as they got.
Today, the report may run on OS/ 390, the database on Unix, and data changes may flow from Windows NT servers of Web-based applications.
 |
 |
|
|
|
"The conventional concept of batch job scheduling is moving from a time-driven process to an event-driven one," wrote Paul Mason, an analyst at Framingham, Mass.-based IDC, in his December report, "Event-Driven Scheduling—the Next Step in System Automation?"
With businesses running globally around the clock, a credit card company's request for a report on a delinquent account can come into Narex at 2 a.m. as easily as at noon, says Paul Konkel, IT director at Narex.
In the old, time-driven model, Narex's systems could be set to automatically begin such jobs. But monitoring and restarting failed jobs was a manual process.
Now, the trigger for the job scheduler to start the report is receipt of the job request. And if the job should abend, or fail, SysAdmiral can look into the application and often detect why, says Konkel. If an address change was not affected, SysAdmiral can detect the problem, trigger the application to update the database and then rerun the job, all without manual intervention.
Having scanned the logs, SysAdmiral can detail conditions under which the process failed—"a big timesaver when you're trying to figure out what went wrong," says Konkel. Reports that once took 24 hours now take Narex eight to 12 hours. Two years after installing Mountain View, Calif.-based Tidal's event-driven job scheduler, that time is still shrinking, Konkel says.
"A job scheduler is one product where you can see a lot of gains from a single, simple tool," Dryden says. "You don't have to buy into a whole framework, and there's an immediate, huge return on investment."
The Wisconsin Department of Corrections was using another product when Linda Johnson took over as applications support engineer. "It's a very complicated product, more than we needed to run 30 jobs a day," she says.
She instead selected Argent Job Scheduler. Farmington, Conn.-based Argent's product is smaller and simpler than what the department had been using but offers the flexibility and reliability Johnson's predominantly Windows NT systems need, as well as compatibility with the state's mainframe and other state agency systems and applications. And support is excellent, Johnson says, an important consideration in a shop with a small staff.
Argent Job Scheduler automates database updates for the state's sex offender Web site, which is now available only to police but soon will be accessible by the public.
"It extracts data from our legacy system and moves that data to the mainframe," Johnson says. Offenders move an average of 2.5 times per year, so just tracking addresses is a big job, she says.
Sophisticated event-driven job schedulers are blurring the line between business-process automation and job schedulers, Mason writes.
Smartworks.com Inc., a Web-based printing company in Dayton, Ohio, uses a job scheduler from Tidal to run jobs around the clock for its Web and SQL applications.
"Tidal is scheduled to kick off internal applications at Smartworks and, in the future, our disaster recovery site," says Eric Gephart, the company's senior network recovery manager. "The next step will be for the Tidal application to run on a server; a SQL back-end [server] will be running on our storage-area network."
When Dan McFarland took over in 1999 as the CIO for the city of Dallas, an early initiative was to install Houston-based BMC's Control-M.
"Now, when a job runs, if it abends, it has autorecovery built in," explains McFarland. "We don't have to call a programmer—Control-M restarts it."
The job scheduler underpins the city's new 311 system, which lets residents file complaints on everything from potholes to real estate assessments. The integrated system uses layers of applications—including databases, reporting programs, geographic information systems and call centers—integrates the information, and lets residents and city officials track complaints via the Web.
It also runs jobs for Dallas' customer relationship management system and will take on similar tasks for a new enterprise resource planning (ERP) system, McFarland says. "Scalability was an important issue for us," he says.
"The importance of job scheduling is very much tied to the ERP space," Mason says.
When Honeywell Aircraft Landing Systems migrated from an OS/390-based ERP system to Glovia, a Unix-based system from Fujitsu Ltd. in Tokyo, the need for a job scheduler external to Glovia was quickly apparent to David Kulakowski, Honeywell's applications development manager.
"We didn't know until the next morning if our applications had failed the previous evening," he says.
South Bend, Ind.-based Honeywell repairs and overhauls commercial aircraft brakes. Glovia tracks and manages the servicing procedures done on wheels and brakes at repair facilities, Kulakowski says. "We have to know exactly what's been done to every part of an aircraft at all times - that's FAA requirements. If we don't, we can be fined," he says.
Kulakowski brought in SysAdmiral to manage the process flows between Glovia and applications at Honeywell's corporate data center in Tempe, Ariz. SysAdmiral does the job with information flowing bidirectionally, he says. "Locally, we don't man our IT systems in the evening now," says Kulakowski.
XML-Enabled
XML capabilities for job schedulers are also growing in importance. XML capability wasn't the first of the Tidal scheduler's features that Konkel implemented at Narex, but it's a popular one among the company's programmers.
"I can have a team of programmers design a process flow in Visio, say, then output it in XML and use it in the Tidal job scheduler," Konkel says. XML allows the process owners to have control over the details of how their jobs run and frees Konkel's staff from having to recode the process.
"Enterprises are taking a more pragmatic perspective of their IT organizations," Dryden says. "Their expectations of the business value that IT should deliver are higher." And delivering that reliability starts with operations, he says.
Ironically, it's the old mainframe job schedulers' success at streamlining operations that can be the greatest obstacle to buying a new one.
"Even if a new and better one comes along, you've got so many rules in place with the old one, it's nearly impossible to tear it out," says Ray Lefebvre, Oracle database administrator at Stride Rite Inc. in Lexington, Mass.
And substituting one scheduler for another isn't always necessary.
Reuters Group PLC in London is extending its IBM and Tivoli Systems Inc. job schedulers across its distributed systems environments to "hundreds of Solaris and NT boxes," says Andrew Cunningham, manager of global management systems at Reuters.
With widely distributed sites, "we needed a tool that would work across multiple environments but that could be controlled centrally," Cunningham says. Because Reuters' systems run under Tivoli Enterprise Console, the decision to use Austin, Texas-based Tivoli's Unix and NT scheduler was a logical one, he says.
"The quest for enterprises to move to a single cross-platform job scheduler remains a key requirement," wrote Milind Govekar, an analyst at Gartner Inc. in Stamford, Conn., in his July report, "Job-Scheduling Magic Quadrant: New Challenges." But for most firms, that's unlikely to happen before 2004.
Lais is a freelance writer in Takoma Park, Md. Contact her at sami_lais@computerworld.com.
Related stories:
- Upsides and Downsides, Feb. 25, 2002
- BMC's software plans stir controversy, June 7, 2001