// the find
man-group/arctic
High performance datastore for time series and tick data
Arctic is a Python time series store built on MongoDB, designed for quant finance workflows — storing pandas DataFrames, numpy arrays, and tick data with versioning and fast compression. The repo is now in maintenance mode; Man Group moved active development to ArcticDB, a C++ rewrite with a columnar storage engine. If you're evaluating this for new work, you want ArcticDB, not this.
The versioned store is the standout feature — you get Git-like snapshots of your DataFrames without any extra tooling, which is genuinely useful for backtesting where you need reproducibility. The tickstore handles high-frequency time series efficiently, chunking by date and compressing with LZ4 so reads stay fast even at scale. The bitemporal store (tracking both event time and knowledge time) is a real feature that most competing tools don't bother with. Test coverage is thorough, with both unit and integration layers clearly separated.
This is dead software — the README says so in the first sentence. Last commit was April 2024 and no new features are coming. It's MongoDB-backed, which means you're pulling in a document store to handle something columnar storage handles better; ArcticDB dropped this dependency entirely for good reason. The async support looks bolted on rather than designed in. Anyone adopting this today is signing up to eventually migrate to ArcticDB anyway, so the only reason to start here is if you have an existing Arctic deployment you haven't migrated yet.