Any plan for publishing job DAG information?

[uchuhimo]

The scheduler may use DAG information in the job to achieve better resource utilization, which has been explored in Graphene. Publishing job DAG information will inspire further exploration to use DAG information in scheduling.

[furykerry]

We are working on it, but it may take some time. You can detail the information you need for DAG.

[uchuhimo]

The most critical DAG information is dependencies between tasks, which indicates the enforced schedule order of tasks. DAG information table may contain two fields: dependents and dependency type. For example, if both a and c depends on b with a many-to-one relationship (b has multiple instances, a has single instance, the instance of a depends on all instances of b), dependents of b is a and c, and dependency type is many-to-one, the corresponding row in DAG information CSV file is b, a|c, many-to-one.

[HaiyangDING]

DAG info is added in the v2018 trace. Enjoy :)

Closing the issue.

[ZZzz1314520]

Hello! I am a student from Chongqing University. I used this data in my graduation project, but I still have some questions about this data and would like to consult with you. Regarding the start_timestamp column and end_timestamp column in the Instance table, there is a task with start_timestamp of 399 and end_timestamp of 401. Does this mean that the task only ran for 2 seconds before it ended? 2 seconds seems a bit short, so I would like to ask if the 2 seconds here needs to be multiplied by a certain value? Have a nice day!