Domino, a Platform-as-a-Service for data science, enables people to do analytical work using languages such as Python or R in the cloud.
According to Nick Elprin (co-founder of Domino), Domino allows data scientist to focus on their analysis, not their infrastructure:
As data volumes have increased and analytical techniques have become more sophisticated, we believe the tools required to do modern data analysis have lagged in their ease of use and have unnecessarily restricted access to the field of data science
The Domino platform rests on three pillars of functionality:
- Direct to cloud deployment and execution: Domino allows to run existing code (Python, R, Matlab, Julia, shell scripts, etc.) on EC2 in order to off-load long-running or resource-intensive tasks. The system also takes care of all the plumbing under the hood to make this happen: AMI management, starting and stopping machines, securely transferring data onto a machine and securely transferring results back.
- Version control for data science: The Domino folks figured that tools like git are insufficient for analytical workflows, as they can’t handle large data sets, and don’t create a link between the inputs and the results (e.g., charts, figures). Domino automatically keeps snapshots of the entire project, currently up to 40GB, making it easy to trace the history of the work in its entirety, incl. code, data, and the results.
- Collaboration: Like a Github project, Domino projects can have collaborators who can view, edit, and run a project. Domino detects conflicts, sends notifications with updated results after runs finish, and has an internal notebook to facilitate discussion as the team’s work progresses.
With a pay-as-you-go approach, Domino’s pricing ranges from introductory free accounts to monthly subscriptions. As InfoQ learned from Nick Elprin, despite its early days the platform is already being used by the entire spectrum of data science practitioners: from academics, such as an ecologist who analyzes thousands of images for her research, over data science consultants for Kaggle competitions to marketing firms, for example, to help its clients better target mailings.
Ralph Winzinger Nov 25, 2014
John Krewson, Steve Ropa and Matt Badgley Nov 24, 2014