Loading the full bookkeeping of versions in memory can be time and resources consuming. This slows down accepting and processing public API endpoints such as transaction, queries, subscriptions, etc. and can create odd-looking behaviour.
This PR tries to spawn as many tasks as possible (all the ones that don't need access to the full bookkeeping), seed Bookie and then launch the tasks that need complete in-memory bookkeeping.
This should significantly improve startup times. It should be instantaneous for nodes that don't have any local versions.
Loading the full bookkeeping of versions in memory can be time and resources consuming. This slows down accepting and processing public API endpoints such as transaction, queries, subscriptions, etc. and can create odd-looking behaviour.
This PR tries to spawn as many tasks as possible (all the ones that don't need access to the full bookkeeping), seed
Bookieand then launch the tasks that need complete in-memory bookkeeping.This should significantly improve startup times. It should be instantaneous for nodes that don't have any local versions.