Batch Jobs Need More CLASS? Group Workloads by Service Groups
From the days programmers began enjoying the luxury of multi-tasking computers, they’ve relied on job classes to define the way they want the system to handle and prioritize workloads. But there is often a disconnect between how programmers assign jobs to classes and how the business sees the importance of various workloads. We need a little more class.
Here is how IBM explains what job classes are used for:
“Use the CLASS parameter if your company uses classes to group jobs. Grouping jobs helps to:
- “Achieve a balance between different types of jobs. A good balance of job class assignments helps to make the most efficient use possible of the system.
- “Avoid contention between jobs that use the same resources.”
While this is a great way to balance resource utilization and ensure that jobs proceed smoothly through the system, it doesn’t always give the results users hope for. To really achieve a well-running, efficient system and happy customers, you need more. Let’s ask for something different for our batch workloads in the JESPLEX.
Grouping Batch Workloads by Service Groups
Instead of simply grouping batch workloads by resource demands, let’s group it by business unit or group. We’ll call these service groups.
Each service group defines the batch importance of the workload. Not only does it affect the execution phase, it also will determine the order in initiation and submission. And you also get to specify queue service times to set goals for how fast a job should be initiated. If the goal is missed, the priority increases to ensure no job is left behind.
At the same time, you want to be sure that you don’t adversely key online work by taking up too many resources managing batch. So, you want a system that will do the best to optimize the batch workloads while understanding and complying with datacenter-defined constraints. You also want the flexibility to set up a variety of policies to deal with weekend, end-of-month and other special times. And if you’re reaching for the moon, go all the way and ensure you can set additional constraints to avoid breaching the CPU capacity limits for your rolling four-hour average.
Service Groups with ThruPut Manager
It turns out the “moon” is accessible. Compuware’s ThruPut Manager does all the above and more. You can literally set and forget your batch by implementing service groups as part of setting up ThruPut Manager. Though the exercise takes a little upfront effort, once you have created a reasonable set of values for your workload groups, you can let that part of the work go. Here’s a simple guide to what you need to do:
While it seems a little complicated, ThruPut Manger’s documentation will walk you through it, and the interface is straightforward. Fairly quickly, you will be enjoying the benefit of a better-managed batch system.
Latest posts by Denise Kalm (see all)
- Breaking Down Dev and Ops Silos with Communication, Collaboration and Trust - April 19, 2018
- What Do Performance and Capacity Roles Have to Do with DevOps? - February 1, 2018
- How Developers Can Boost DevOps Through Automated Batch Processing - December 7, 2017