WeatherXM Demo for CoD Summit @ LabWeek22

Compute on Weather Data using Compute over Data with Bacalhau!

Prerequisites

• Docker
• Bacalhau cli
• Datasets of weather data from WeatherXM weather stations - both observations and forecasts

Code

A simple script is provided that caclulates statistical indexes, produces helpful images for the researcher and lastly prints out the results. It expects exactly two command line arguments:

• the path to the directory with the weather data
• the path to the directory where the images should be stored

Running

All you need to do to run the workload on Bacalhau is to provide

1. the CID of the folder with your datasets
2. the public URI of your IMAGE
3. the paths of the INPUT_DIR and OUTPUT_DIR directories

Then by running bacalhau docker run -v ${CID}:/inputs ${IMAGE} python -- QoF.py ${INPUT_DIR} ${OUTOUT_DIR} your workload will be submitted for execution to the network. An example would be:

bacalhau docker run -v QmVc7o6daLsVoQYUoFmp3rkD8XXD9pXnmsbgyTWmB89w3h:/inputs ghcr.io/weatherxm/cod-demo:1.0.0 python -- QoF.py /inputs/ws_lisbon /outputs

To take advantage of the sharding capabilities you can use a glob pattern instead of a specific path as INPUT_DIR, for example

bacalhau docker run -v QmVc7o6daLsVoQYUoFmp3rkD8XXD9pXnmsbgyTWmB89w3h:/inputs --sharding-base-path "/inputs" --sharding-glob-pattern "ws_*" --sharding-batch-size 1 ghcr.io/weatherxm/cod-demo:1.0.0 python -- QoF.py "/inputs/ws_*" "/outputs"

Developing

If you want to experiment with the datasets but want to alter the code, all you have to do is to provide your custom image. When the code is ready you can build a new image and TAG it with docker build -t ${TAG} . and later push it to a public image repository (eg docker-hub)