services/horizon: Benchmark reingestion based on precomputed ledger meta tx files
Opened this issue · 1 comments
sreuland commented
What problem does your feature solve?
- absence of performance benchmark numbers for time it takes to reingest at least 1 year of latest pubnet data.
- public docs haven't been updated for new reingestion from datastore capability.
What would you like to see?
-
A dedicated runtime environment with at least provides the horizon required compute resources
-
Run
db reingest range --parallel-workers <?> --ledgerbackend datastore --datastore-config config.toml <1 year ago> <current>
-
Monitor the performance of cpu/memory, runtime duration.
-
Try different performance config parameters:
- parallel workers: likely 5 or less.
- buffered storage config per results from #5390
-
Update the public docs for reingestion to state this new
--ledgerbackend datastore
option.
What alternatives are there?
sreuland commented
@urvisavla , should we move this ticket up to sprint 49 backlog?