-
Notifications
You must be signed in to change notification settings - Fork 205
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add LaunchMultiTask rpc interface for executor #209
Comments
@yahoNanJing I looked into this and just wanted to clarify something. I might be wrong, but looks like the current proto definition allows to send multiple tasks in a single launch task. See line However in the code, I see the tasks are launched one at a time. See line Can this task be simplified as modifying the |
Hi @askoa, yes currently we can launch a bunch of tasks in a single grpc request. However, for the same stage, they don't share the information, like execution plan, etc. I'd like to add something like this.
Then we can not only reduce the rpc request body size, but also reduce the execution plan deserialization cost. |
@yahoNanJing Could you please check the protobuf changes in the draft PR and let me know if it looks okay? |
Thanks @askoa. Left some comment on your draft. Are you planning to implement this or already have an implementation for this? Actually, we have already implemented a version for this and plan to contribute a patch for it. |
@yahoNanJing I just started with protobuf changes and not implemented yet. I'll drop the draft PR. Thanks for letting me know. |
Is your feature request related to a problem or challenge? Please describe what you are trying to do.
Sometimes, the scheduler will schedule multiple tasks of the same stage to the same executors. Currently we have to create a task definition for each task, which may not be efficient.
Describe the solution you'd like
For the above case, it's better to provide a LaunchMultiTask rpc interface for the executor so that multiple tasks of the same stage can be just one MultiTaskDefinition to avoid the same plan be copied multiple times. The network io cost can be reduced.
Describe alternatives you've considered
Additional context
The text was updated successfully, but these errors were encountered: