mingmwang opened a new issue, #116:
URL: https://github.com/apache/arrow-ballista/issues/116

   **Is your feature request related to a problem or challenge? Please describe 
what you are trying to do.**
   A clear and concise description of what the problem is. Ex. I'm always 
frustrated when [...] 
   (This section helps Arrow developers understand the context and *why* for 
this feature, in addition to  the *what*)
   
   In the current Ballista implementation, when the Executor finish the tasks, 
only the task status(Failed/Complete) is report back to the Scheduler, the plan 
metrics are not collected and further combined in the Scheduler, the Scheduler 
does not have an overall view of plan execution metrics, like how many rows are 
scanned, total elapse times of each operator etc.
   
   **Describe the solution you'd like**
   A clear and concise description of what you want to happen.
   
   1. In the Executor side, when the task is finished, need to collect all the 
plan operators metrics for that task.
   2. Send the plan metrics back to Scheduler along with TaskStatus.
   3. Scheduler need to further combine/aggregate the metrics from all the 
completed tasks and finally update the plan metrics to
   the physical plan hold by the Scheduler. 
   4. It is better that DataFusion can add an unique id to 
ExecutionPlanMetricsSet, so that Ballista Scheduler can easily combine and 
aggregate metrics.
   5. The metrics combine/aggregation overhead could be quite expensive, 
considering there might be tens of thousands of tasks for a heavy stage, the 
metrics aggregation should be handled in an async way and should not impact the 
major stage/task scheduling and task result processing flow and allow dropping 
the metrics if the metrics combine/aggregate speed can not catch up.
   
   
   **Describe alternatives you've considered**
   A clear and concise description of any alternative solutions or features 
you've considered.
   
   **Additional context**
   Add any other context or screenshots about the feature request here.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to