Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
0.23.0, 1.0.0
-
None
-
None
Description
Currently gridmix tries to gauge cluster load etc. and throttles job submission. This makes it unpredictable and also is hard to support across MR1 and MR2.
I propose we simplify it to be:
- Replay mode - Just submit jobs in the interval as in the original trace.
- Stress mode - Compress the interval with a given factor for all jobs.