Police EIS Computer System Requirements

Home/Projects/Public Safety/Preventing Adverse Police Incidents/Police EIS Computer System Requirements

There are several computational considerations for police departments that would like to use our EIS. This page addresses the most common ones.

Server Specifications

The server specifications required to store, process, and analyze your data depend on how much you have. Large departments that have a lot of historical data (more than six years) heavily probably need a few terabytes. Medium departments that have a medium amount of historical data (five to six years) probably need hundreds of gigabytes. Small departments may only need tens of gigabytes.

Modeling can be computationally expensive, but you do not need to run models all the time. If possible, we recommend you keep a small machine to generate and explore predictions (perhaps a couple cores and 4-8 GB of RAM) from existing models and using large, temporary machines to handle modeling tasks. If temporary scaling is a challenge, most departments can do well with 2-4 cores and 16-32 GB of RAM.

Software Specifications

We use all free software, including Python (packages listed here), Linux command-line tools (e.g. drake), and Postgres.

Why Do We Recommend Linux?

There are a couple basic reasons for going with Linux for this project:

  1. It’s faster, easier, and cheaper to deploy the system on Linux. The pipeline was written on Linux and uses Linux-specific tools. Those would need to be modified to implement the system in a Windows environment. Some of the software that is supposed to be OS-independent seems to run into trouble. It would take time and effort (and hence money) to fix those problems.
  2. The pipeline will improve, and the changes will be made in Linux. We use Linux, and we continue to improve the EIS and build tools that help our partners. (For example, we’re writing a piece of code that automatically creates thousands of potential predictors, which improves accuracy, and another piece of code that displays model evaluations in a webapp.) It’s easier to use our work if you use Linux. You would either have to fix linux-Windows incompatibilities to use those or forego free updates.

    In addition, our partners are adopting Linux for this project and have agreed to share useful code. You may miss their work if you pass on Linux.