piker
piker copied to clipboard
Caching subsys: db, filesystem, do we even need something?
Recently there's been some discussion on a caching layer for stuff like:
-
MktInfodata that will land in #489 - general provider symbology (schemas) and dynamic update of venue instrument offerings
- accounting state which is currently entirely managed in
TOMLfiles on the local filesystem of thebrokerdactor. - aggregated multi-trader order dialog states
maybe more?
FOSS systems worth checking out in this space:
- we already are using
elasticsearchfor some option OI tracking -
rocksdb: https://rocksdb.org/ (from META) - repo: https://github.com/facebook/rocksdb/
- pyclient: https://github.com/stephan-hof/pyrocksdb