According to leading industry analysts, batch processing accounts for about 50 percent of the work done in most data centers. Although much of that work is performed by mainframes, they share that role with a diverse mix of distributed servers. The result is a heterogeneous IT landscape. These complex environments demand an enterprise-grade management solution such as one provided by enterprise job schedulers that offer breadth of platform and application coverage.
While the marketplace offers enterprise scheduling solutions that address these issues to varying degrees, until recently, they all shared a common characteristic that limited their scope of deployment and, therefore, their overall effectiveness. To schedule jobs on a particular platform, an agent had to be deployed, causing a significant impact on the Total Cost of Ownership (TCO ). Some of the TCO considerations are the time and effort required to initially install and service each agent. Each new vendor release requires upgrading, and if a new server is commissioned, the installation and maintenance must be re-applied.
The list of issues extends beyond simple installation, as the agent deployed must communicate with a server or manager component. Typical communication is implemented using TCP /IP. If an agent is required outside the immediate network segment and a firewall or similar device is involved, ports may need to be opened to accommodate the agent-to-server connection. If an organization has stringent security policies, the deployment of the scheduling agent may be delayed or denied if sufficient justification can’t be made. Furthermore, scheduling agents must be able to run jobs on behalf of a wide variety of users. This requires the agent to operate in super-user or administrator mode. This, too, may contradict security policies and even be considered a breach of corporate security.
Finally, agents are built and certified for specific platforms and operating system versions. Organizations that wish to or must deploy leading edge operating systems may find their scheduling vendor hasn’t yet upgraded or certified their scheduling for that platform. This may cause project delays.
The solution is emerging—agent-less scheduling. Just as we’ve seen in other disciplines, such as network management, server management and even identity management, job or workload management is entering the agent-less era. The potential benefits of agent-less batch job scheduling are vast and all the issues discussed prior are addressed by this approach. The immediate benefit is a substantial reduction in the total cost of managing an enterprise job scheduling solution. A typical Global 2000 organization manages more than 1,000 servers. Thus, the savings of bypassing installation, maintenance and administration of even 50 percent of those servers can be sizable. Once enterprise job scheduling is freed from the constraints of agent installation, the reach and scope of the enterprise solution can be fully leveraged throughout the organization to, for example, ensure file backups are performed as they should be and logs are collected to adhere to legislated and organization policies. Support for change management and problem resolution takes advantage of the automation capabilities of a job scheduler, freeing analysts and IT staff from having to manually reach out to servers that previously haven’t participated in the enterprise batch management “domain.”
Because agent-less technology uses standard and wellknown ports, firewall configuration and security management cease to be impediments to extending the benefits of enterprise scheduling to external servers.
Agent-less job scheduling will usher in a new era of enterprise flexibility at a far lower cost of operation and ownership than past agent-based solutions. The true business benefit of the agent-less technology is that now companies can install batch management capabilities on more machines than ever before—allowing them to see more information and make better informed business decisions. This brave new world has been made possible by the fact that distributed systems now embed either Secure Shell (SSH ) or Windows Management Instrumentation (WMI) technology. It’s somewhat ironic that the mainframe has had agent-less job scheduling for more than 25 years with Network Job Entry (NJE). Once again, the mainframe leads with technology that has taken distributed systems a while to finally adopt. Z