Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Terminology

Here are some terms about scheduled tasks in Apache Griffin

  • Measure, an entity which determines a process logic on certain properties of data.
  • Job, an entity which represents a specific task under the evaluation rule of a Measure.
  • JobSchedule, an entity which determines how to schedule a Job
  • Metric, an entity which describes the computation results of a Job

Relationship

A measure can map many jobs, and each job corresponds to a metric.

Image Added

Schedule Process

Steps to schedule a job and get the evaluate result:

  1. User defines a JobSchedule entity, which maps a certain Measure and contains the rule about what time to execute the Job and what part of data to use.
  2. Griffin creates a Job entity corresponds to the JobSchedule, and Griffin Job Scheduler performs the computation according to the settings in the JobSchedule.
  3. Every scheduled job instance produces a result, and it is stored as an entity of MetricValue.
  4. Entities of MetricValue which come from the same Job form a Metric entity, and user can use the entities of Metric to do analysis and so on.

Image Added

 

More details about Griffin Job Scheduler can be found here:

https://cwiki.apache.org/confluence/display/GRIFFIN/5.+Griffin+Job+Scheduler+Design

Image Removed

  1. User defines a measure, setting an organization and some data assets with it, which means one measure mapping to one organization, and one or two data assets.
  2. User also defines a schedule, setting schedule information, corresponding nothing firstly.
  3. User chooses a measure and a schedule, creates a job, which describes how and when to execute a measure task. One job mapping to one measure and one schedule.
  4. One job needs to be executed periodically in griffin measure module, which means one job mapping to one metric.
  5. Griffin measure module generates metric data periodically, the metric data with the same metric id compose the metric. One metric mapping to many metric datas.