by Carol Baroudi, Partner
Gone are the days of simple batch processing. Welcome to the oh-so-very-flat world that is embracing service-oriented architectures, grid architectures, expansive ecosystems, and end-to-end processing. What once was complicated enough just got a heck of a lot more complicated.
As soon as multiple events need to happen at the same time, some precedence, some orchestration has to happen. Whether it’s a traffic light or air traffic control or workload scheduling, the consequences of inadequate management are predictably disastrous. The concept of scheduling is not new to IT. Operating systems schedule, routers schedule, and traditional job scheduling programs schedule. However, the multi-dimensional aspects of contemporary IT require job scheduling to take a quantum leap. Here’s why:
Introducing Service Oriented Architecture – A job is not a job is not a job
Enterprises have begun to recognize the rigidity and inefficiency of traditional applications silos and have begun movement toward service-oriented architecture. Traditional applications have a beginning, a middle and an end, and lend themselves very well to traditional job scheduling that can create and prioritize a job queue that understands when a job starts and when it stops. In the world of SOA, traditional applications are deconstructed, their core functionality exposed as services that recombine in countless ways in numerous composite applications. The concept of a discrete job begs redefinition. And the idea that a particular service, instantiated in multiple composite applications, could be subject to simultaneously different service level requirements is enough make an engine seize.
Introducing Grid Architecture and Virtualization – A resource is not a resource is not a resource
Just as SOA wreaks havoc with the concept of job, grid architectures and virtualization wreak havoc with the concept of resource. The jobs that need to be scheduled and the resources that need to be scheduled are two critical components of scheduling. Grid architectures pool resources to deliver capacity beyond that available from individual resources. Virtualization redefines, reallocates, and simulates resources. The concept of a resource being a fixed entity of known capacity, available or not available, is lost in both the world of virtualization and in grid architectures. Traditional job scheduling doesn’t take the newer definition of resource into account.
It’s time for workload scheduling to take a quantum leap. The folks at CA (Computer Associates) are paying close attention and have begun to execute against a workload automation roadmap that targets Service Level Agreement (SLA)-based automation for 2008 and supports grid by 2009. Hurwitz & Associates believes it’s not a moment too soon.