Main repository for the NeuroCAAS project, providing a neuroscience analysis platform using Infrastructure-as-Code (IaC).
This repository hosts all of the Code that makes up our IaC approach, with structure as described in the paper.
Please note that this project is released with a Contributor Code of Conduct, here.
Structure:
- paper (material specifically related to the NeuroCAAS paper)
- docs (documentation and developer guides)
- media (demo movies, slides, etc.)
- experiments (benchmarking code for NeuroCAAS on existing algorithms)
- ncap_iac (the core code of the project)
- ncap blueprints (code representing analysis pipelines.)
- user profiles (code representing users.)
- protocols (code representing job management coordination analysis pipelines and users.)
- utils (utilities to compile the above code into the corresponding infrastructure. Interfaces with troposphere and the AWS CLI)
See the developer guide.
You will need to install the dependencies found in the requirements_experiments.txt file in order to run experiments and compare NeuroCAAS to your own infrastructure. We recommend doing so in a conda virtual environment. Once you have installed conda, check your installation by running:
conda list
Then create a new environment as follows:
conda create -n neurocaas_experiments python=3.6
Now clone and move into the root directory of this repo.
git clone https://github.com/cunningham-lab/neurocaas.git
cd /path/to/this/repo
Activate your new environment, and install necessary packages by running:
conda activate neurocaas_experiments
conda install pip
pip install -r requirements_experiments.txt
Navigate to the experiments directory, and run the following to generate png files shown in Figure 4:
python recreate_figure4.py
Pngs showing cost and time analyses will be stored in the "panels" subdirectory within each analysis. The files suffixed "Cost", "Time", "LCC_powermatch", and "LUC_powermatch" are those shown in the submitted figure. The source code to generate these analyses and figures can be found in the "calculate_cost" module, called by the "getdata_fig4" function.
You can generate the data for custom versions of Figure 4 by filling out a custom cost file. See experiments/Custom_CostFiles/hardwarecost.yaml and experiments/Custom_CostFiles/hardwarecost2.yaml for annotated examples. Once you have filled out a cost file, you can run it by executing the following commands:
cd /path/to/local/repo/neurocaas/experiments
python generate_customcomparison.py /path/to/your/costfile.yaml
Developers can calculate analysis usage and parallelism for their own analyses once deployed on NeuroCAAS, as shown in Figure 6. First install the neurocaas contrib package. Instructions found at https://github.com/cunningham-lab/neurocaas_contrib. Then generate log files for your analysis, to be written to some folder:
neurocaas-contrib init ## enter the name of your analysis
neurocaas-contrib monitor visualize-parallism -p /path/to/logfolder
Running this command will generate JSON files that describe individual jobs run by individual users in analysis buckets as:
{analysis_name}_{user_name}_{job_timestamp}_parallel_logs.json, where each log contains individual jobs.
Then, use the script neurocaas_contrib/figures/parallelized.py to create an analogue to Figure 6:
python /path/to/neurocaas_contrib/figures/parallelized.py /path/to/logfolder
To generate a figure in that same folder, parallelism_figure_{date time}.png. If you have multiple analyses, you can generate the logs for all of them into the same log folder, and the script parallelized.py will create a graph that includes all of them.
Note that the script parallelized.py includes conditional statements to exclude debugging jobs run in the course of creating NeuroCAAS infrastructure as well.
To run our comparison analysis on your own infrastructure, navigate to the "experiments" directory and follow the directions found in the guide there.
See the roadmap for expected timeline of new analyses, new features, and supplementary packages for this project.