/xorbits

Scalable Python data science, in an API compatible & lightning fast way.

Primary LanguagePythonApache License 2.0Apache-2.0


Xorbits: scalable Python data science, familiar & fast.

PyPI Latest Release License Coverage Build Status Doc Slack Twitter

What is it?

Xorbits is a scalable Python data science framework that aims to scale the whole Python data science world, including numpy, pandas, scikit-learn and many other libraries. It can leverage multi cores or GPUs to accelerate computation on a single machine, or scale out up to thousands of machines to support processing terabytes of data. In our benchmark test, Xorbits is the fastest framework among the most popular distributed data science frameworks.

As for the name of xorbits, it has many meanings, you can treat it as X-or-bits or X-orbits or xor-bits, just have fun to comprehend it in your own way.

At a glance


Codes are almost identical except for the import, replace import pandas with import xorbits.pandas will just work, so does numpy and so forth.

Where to get it

The source code is currently hosted on GitHub at: https://github.com/xprobe-inc/xorbits

Binary installers for the latest released version are available at the Python Package Index (PyPI).

# PyPI
pip install xorbits

API compatibility

As long as you know how to use numpy, pandas and so forth, you would probably know how to use Xorbits.

All Xorbits APIs implemented or planned include:

API Implemented version or plan
xorbits.pandas v0.1.0
xorbits.numpy v0.1.0
xorbits.sklearn Planned in the near future
xorbits.xgboost Planned in the near future
xorbits.lightgbm Planned in the near future
xorbits.xarray Planned in the future
xorbits.scipy Planned in the future
xorbits.statsmodels Planned in the future

Lightning fast speed

Xorbits is the fastest compared to other popular frameworks according to our benchmark tests.

We did benchmarks for TPC-H at scale factor 100 (~100 GB datasets) and 1000 (~1 TB datasets). The performances are shown as below.

TPC-H SF100: Xorbits vs Dask

Xorbits vs Dask

Q21 was excluded since Dask ran out of memory. Across all queries, Xorbits was found to be 7.3x faster than Dask.

TPC-H SF100: Xorbits vs Pandas API on Spark

Xorbits vs Pandas API on Spark

Across all queries, the two systems have roughly similar performance, but Xorbits provided much better API compatibility. Pandas API on Spark failed on Q1, Q4, Q7, Q21, and ran out of memory on Q20.

TPC-H SF100: Xorbits vs Modin

Xorbits vs Modin

Although Modin ran out of memory for most of the queries that involve heavy data shuffles, making the performance difference less obvious, Xorbits was still found to be 3.2x faster than Modin.

TPC-H SF1000: Xorbits

Xorbits

Although Xorbits is able to pass all the queries in a row, Dask, Pandas API on Spark and Modin failed on most of the queries. Thus, we are not able to compare the performance difference now, and we plan to try again later.

For more information, see performance benchmarks.

Deployment

Xorbits can be deployed on your local machine, largely deployed to a cluster via command lines, or deploy via Kubernetes and clouds.

Deployment Description
Local Run Xorbits on a local machine, e.g. your laptop
Cluster Deploy Xorbits to existing cluster via command lines
Kubernetes Deploy Xorbits to existing k8s cluster via python code
Cloud Deploy Xorbits to various cloud platforms via python code
SLURM Deploy Xorbits to SLURM

License

Apache 2

Documentation

The official documentation is hosted on: https://doc.xorbits.io

Roadmaps

Main goals we want to achieve in the future include:

  • Transitioning from pandas native to arrow native for data storage,
    it will reduce the memory cost substantially and is more friendly for compute engine.
  • Introducing native engines that leverage technologies like vectorization and codegen to accelerate computations.
  • Scale as many libraries and algorithms as possible!

More detailed roadmaps will be revealed soon, stay tuned!

Relationship with Mars

The creators of Xorbits are mainly those of Mars, we built Xorbits currently on Mars to reduce duplicated work, but the vision of Xorbits suggests that it's not appropriate to put everything into Mars, instead, we need a new project to support the roadmaps better. In the future, we will replace some core internal components with other upcoming ones we will propose, stay tuned!

Getting involved

Platform Purpose
Discourse Forum Asking usage questions and discussing development.
Github Issues Reporting bugs and filing feature requests.
Slack Collaborating with other Xorbits users.
StackOverflow Asking questions about how to use Xorbits.
Twitter Staying up-to-date on new features.