Multi-week schedules take many seconds to generate.
Observations
Right now, partitioning is done on a per-target basis. However, we can further optimize scheduling and reduce computation time by sharding inputs even more.
Action Items
[ ] Make a load balancer rpc that uses the same protobuf definitions as the Scheduler#schedule rpc
[ ] Partition targets by window of time slots.
[ ] On each retrieval of the async_stream! on the client side, send an event notifier to the frontend to display update progress in webview/src/lib/modals/Processing.svelte
[ ] Run static evaluation on compiled MCTS results, post load balance pass.
mrodz
commented
1 month ago
Problem
Multi-week schedules take many seconds to generate.
Observations
Right now, partitioning is done on a per-target basis. However, we can further optimize scheduling and reduce computation time by sharding inputs even more.
Action Items
protobufdefinitions as theScheduler#schedulerpcasync_stream!on the client side, send an event notifier to the frontend to display update progress in webview/src/lib/modals/Processing.svelte