In-Depth
Job Scheduling: Product Capabilities and Market Trends
In the second part of our three-part series on workload automation, we compare mainframe- and distributed-based job schedulers.
by Mike Gilbert
This is the second part of our three-part series that compares mainframe-based and distributed-based job schedulers and assesses the readiness of Microsoft Windows-based job scheduling technology to manage workloads at the enterprise level.
Last week we offered a brief outline of the evolution of workload automation technology. This week we provide an analysis of job scheduling capabilities on both mainframes and distributed platforms based on detailed comparisons of a selection of products targeting these platforms.
Analysis of Job Scheduling on Windows Job schedulers employ a central server and a database; these two components govern the core scheduling capabilities. Product vendors have chosen either mainframes or distributed platforms to host the central server component and use agents to reach other platforms. Some vendors provide a single console to coordinate a combination of mainframe and distributed servers. No matter what the configuration, the scheduling servers have platform-specific capabilities that affect overall job scheduling capability.
We have interviewed vendors and gathered detailed product information about a selection of key mainframe-based and distributed-based products, including
BMC CONTROL-M for z/OS, CA Unicenter CA-7 Workload Automation, CA ESP Workload Automation for z/OS, and IBM Tivoli Workload Scheduler for z/OS for the mainframe; and BMC CONTROL-M for Distributed Systems, CA Unicenter AutoSys Workload Automation, CA dSeries Workload Automation, IBM Tivoli Workload Scheduler SMA OpCon/xps, Tidal Enterprise Scheduler, and UC4:global for distributed platforms.This primary research was supplemented by Web research on AppWorx, ASCI ActiveBatch, ASG-OpsCentral, ORSYP Dollar Universe, Redwood Cronacle, and Vinzant Global ECS.
Technology
Feature descriptions for batch processing, workload management, and workload automation were used as the basis for collecting and aggregating the relative strength and maturity on the two platforms. The assessment is limited to scenarios where mainframe-based schedulers are used primarily to schedule mainframe workload, and Windows-based schedulers are used primarily to schedule distributed workload.
Mainframe workloads scheduled by mainframe-based job schedulers run under the control of JES2 or JES3 and are scripted using JCL. JES2 and JES3 are therefore important components of the overall mainframe job scheduling environment.
Service classes: Service classes were built into JES2 and JES3 from the outset, to provide a mechanism for prioritizing batch processing based on resource requirements and job priority. Although similar mechanisms have been built into distributed-based products, service classification has been a lower priority.
JCL restart and recovery: JCL and the JES engines support multi-step processes with a sophisticated checkpoint and restart capability not found in the equivalent scripting languages used on distributed platforms for workload execution.
Distributed platform workloads scheduled by distributed-based job schedulers may be scripted using a proprietary scripting language or by platform-native command shell languages. Because there is no standard scripting language, this summary makes no assumptions about scripting features.
Event-based automation: Distributed-based schedulers offer a richer variety of built-in event types, including storage threshold detection, database events, file arrival/update, Web server events, user logon/logoff, mailbox events and network events.
Graphical tools: All job schedulers use Windows GUI tools or Web browsers to provide a central point of control for schedule definitions and operations. However, you can expect richer tools for workload analysis, graphical workflow definition, forecasting and planning functions in schedulers designed for the Windows platform.
Scheduling Web applications: Job schedulers on the Windows platform provide greater opportunities to integrate J2EE and Microsoft .NET workloads. Some schedulers can schedule Web services.
These relative strengths give a broad indication of the differences you may expect from schedulers running on the two classes of host platform. These differences are slight, and vary by product; they are outweighed by the majority of features that show equivalent strength on mainframe and distributed platforms.
Market Forces
Several forces are at work on job schedulers, including cost, the Internet, compliance, complexity, and agility.
Cost: Mainframe users look for ways to optimize the use of shared IT resources to minimize operational costs and to avoid costs associated with capacity upgrades. Job scheduling provides a means to balance workloads across existing servers and to shift non-essential workloads away from peak on-line processing periods. On distributed platforms, costs are lower, thus reducing the pressure for 100 percent utilization at all times. The dominant concern for many organizations is the cost of integrating and automating complex on-line systems and processes. Workload management and automation address this concern by reducing the need for costly real-time integration services and software and by simplifying operational procedures.
The Internet: This is one of the key driving forces for software innovation The Internet is “always on” so there is little need for overnight batch activity when Web, application, and database servers must be responsive to users around the globe in all time zones. Self-service Web applications create a need for rich real-time integration of back-office applications to fully automate the service that is provided to customers.
Regulatory compliance: To be compliant with new regulations, particularly in the health-care and financial services sectors, IT organizations must ensure that key processes are fully automated, monitored, and logged for future audits. Job schedulers are fulfilling a need for coordinated cross-platform integration between servers, packaged applications, and core applications to ensure that business activities can be traced through the various IT components that support them.
Growing complexity: Systems administration to optimize resource utilization becomes more complex with the increasing use of server farms, blades, clustering, virtualization, and storage area networks, which in turn demands greater sophistication in workload management and automation features. Job scheduling can now be extended to optimizing and improving the use of high-end servers.
Agility: The current drive for “agile” IT means there is high demand for software that can be used to achieve simple but rapid integration across a variety of applications and platforms. Job scheduling vendors have responded to this demand by adapting batch integration techniques for use in real-time integration (event-driven scheduling). The key difference between job schedulers and application integration software for real-time integration is that the former offers a quick way to assemble a business process without the cost and effort associated with software development.
There is a great deal of overlap in the market forces on the mainframe and distributed platforms, but these drivers help to explain how job scheduling has evolved differently on these platforms.
Summary
The arrival and adoption of job schedulers on Windows has brought with it a rich and mature capability for more traditional batch processing and workload management. These job schedulers are modelled on, or have been ported from, traditional mainframe tools and thus exhibit the characteristics required to move large batch workloads to the Windows platform.
The need for workload automation—real-time integration—is driving innovation in the job schedulers that run on distributed and mainframe platforms. There is less innovation in batch processing except where extremely high-workload volumes require the use of multi-system computing architectures to handle high and highly variable capacity demands.
In the final instalment of this three-part series, to appear next week, we examine JCL emulation products and review the status of job scheduling products for the Windows platform.
- - -
Mike Gilbert is an independent consultant and owner of Legacy Directions where he is responsible for providing advisory services to the IT industry focusing on the issues around legacy technology. You can reach Mike at mike.gilbert@legacydirections.co.uk.