- Batch is managed to service goals based on simple wall-clock metrics
- Automatic escalation ensures more important work is favored when resources are stressed
- Batch service is reported to all stakeholders in terms of the goals
- Facilitates creation of Service Level Objectives with users
ThruPut Manager AE adopts a service perspective. Throughout all its functions, it is guided by the user goals for their jobs as well as the datacenter goals (or constraints) for the installation as a whole. A common set of metrics, using straight-forward wall-clock values, are used to express these goals, guide the automation engine and give feedback. Displays and reports use the same metrics to give feedback to users, operations, performance analysts and management. Now everyone can communicate in the same language and mean the same thing.
ThruPut Manager AE and the Batch Service Model:
- The datacenter manager is responsible for taking business goals and datacenter constraints into account and developing a policy to guide Operations in the running of the workload. Until ThruPut Manager AE there was nowhere to store this understanding, much less have the system automatically use it when processing the workload. TM AE collects goals and constraints in a policy. Only one policy is active at a time. There can be a different policy for different situations such as DAYTIME, WEEKNGHT, MONTHEND, HOLIDAY and so on.
- ThruPut Manager AE captures service objectives in terms of queue time and execution time goals, as well as relative importance. These are the metrics the automation engine uses to decide which jobs to favor.
- Technical Support staff understand the complex nature of the datacenter resources. Some jobs have to run on certain machines to comply with software licences, or to have access to the correct DBMS, and so on. At run-time ThruPut Manager AE uses constraint information to automatically distribute the workload.
- ThruPut Manager AE provides an “at-a-glance” view of the entire batch workload such as “batch is being selected within its target and the trend is stable”. When the status is “being selected beyond Acceptable and the trend is deteriorating”, manual intervention is likely still not required since automatic escalation is already doing everything possible.
- ThruPut Manager AE provides extensive feedback on every job. Users can check on its status while executing, and they receive a detailed report explaining the service the job received, including sources of delays.
- Technical Support can verify that each business unit is receiving the service they are expecting and explain any periods when those goals could not be met.