Arctic is a high performance datastore for numeric data. It supports Pandas, numpy arrays and pickled objects out-of-the-box, with pluggable support for other data types and optional versioning.
Arctic can query millions of rows per second per client, achieves ~10x compression on network bandwidth, ~10x compression on disk, and scales to hundreds of millions of rows per second per MongoDB instance.
Arctic has been under active development at Man AHL since 2012.
pip install git+https://github.com/manahl/arctic.git
mongod --dbpath <path/to/db_directory>
from arctic import Arctic
# Connect to Local MONGODB
store = Arctic('localhost')
# Create the library - defaults to VersionStore
store.initialize_library('NASDAQ')
# Access the library
library = store['NASDAQ']
# Load some data - maybe from Quandl
aapl = Quandl.get("NASDAQ/AAPL", authtoken="your token here")
# Store the data in the library
library.write('AAPL', aapl, metadata={'source': 'Quandl'})
# Reading the data
item = library.read('AAPL')
aapl = item.data
metadata = item.metadata
VersionStore supports much more: See the HowTo!
Plugging a custom class in as a library type is straightforward. This example shows how.
Arctic provides namespaced libraries of data. These libraries allow bucketing data by source, user or some other metric (for example frequency: End-Of-Day; Minute Bars; etc.).
Arctic supports multiple data libraries per user. A user (or namespace) maps to a MongoDB database (the granularity of mongo authentication). The library itself is composed of a number of collections within the database. Libraries look like:
- user.EOD
- user.ONEMINUTE
A library is mapped to a Python class. All library databases in MongoDB are prefixed with 'arctic_'
Arctic includes two storage engines:
- VersionStore: a key-value versioned TimeSeries store. It supports:
- Pandas data types (other Python types pickled)
- Multiple versions of each data item. Can easily read previous versions.
- Create point-in-time snapshots across symbols in a library
- Soft quota support
- Hooks for persisting other data types
- Audited writes: API for saving metadata and data before and after a write.
- a wide range of TimeSeries data frequencies: End-Of-Day to Minute bars
- See the HowTo
- TickStore: Column oriented tick database. Supports dynamic fields, chunks aren't versioned. Designed for large continuously ticking data.
Arctic storage implementations are pluggable. VersionStore is the default.
Arctic currently works with:
- Python 2.7
- pymongo >= 3.0
- Pandas
- MongoDB >= 2.4.x
Arctic has been under active development at Man AHL since 2012.
It wouldn't be possible without the work of the AHL Data Engineering Team including:
- Richard Bounds
- James Blackburn
- Vlad Mereuta
- Tom Taylor
- Tope Olukemi
- Drake Siard
- Slavi Marinov
- Wilfred Hughes
- Edward Easton
- ... and many others ...
Contributions welcome!
Arctic is licensed under the GNU LGPL v2.1. A copy of which is included in LICENSE