There are several possible useful hints which can be displayed in UI:
how long it's until the current experiment will be done
a warning if experiment takes unusually long time
when each experiment in the queue will start and how long it will take
when configuring experiments, when it will start and how long it will take
Solution
When there already are successful experiments for the same combination of the benchmark and the system instance, we can use their results to roughly estimate how long an experiment will take. It's okay for some hints to be missing due to nonexisting data, we can just display what's available. Failed and timed out experiments should not be considered.
Later we can also take into account docker image hashes in case there are a lot of experiment data available.
Problems
~The runtime KPI is not defined universally, every benchmark uses its own.~
User story
There are several possible useful hints which can be displayed in UI:
Solution
When there already are successful experiments for the same combination of the benchmark and the system instance, we can use their results to roughly estimate how long an experiment will take. It's okay for some hints to be missing due to nonexisting data, we can just display what's available. Failed and timed out experiments should not be considered.
Later we can also take into account docker image hashes in case there are a lot of experiment data available.
Problems