gautierdelorme/marinewatch

find best navigable routes

Marinewatch

See the slides or the report (french) for details.

Overview

General

Lambda Architecture

Batch processing

Streaming Processing

Developer setup

Requirements

• rbenv
• docker (+ docker-compose)

Project setup

Go to the project directory

git clone git@github.com:gautierdelorme/marinewatch.git
cd marinewatch

Start docker containers

docker-compose up -d

Install the required version of Ruby

rbenv install
rbenv rehash

Install Bundler

gem install bundler
rbenv rehash

Install required gems for the web api

cd ./code/web_api
bundle install
rbenv rehash

Project Overview

Technologies

• Docker as containers manager
• Batch and Streaming processing using Spark and HDFS (written in Scala)
• Neo4j as graph database
• Web and Streaming APIs built in Ruby (using Sinatra)
• CLI tool written in pure bash
• Git as versioning system

Architecture

• code/: contains all the marinewatch code
  • mwspark Spark Scala application
    • Batch processing to generate structured data to be imported in Neo4j
    • Streaming processing to update Neo4j data in real time
  • web_api Web API to get shortest path between two geo coordinates (supported formats: html, json and kml)
  • streaming_api Streaming API to push new updates from boats
• data/: contains all data files
  • input Data files used by Spark jobs
  • output Data files generated by Spark jobs
• neo4j/:
  • conf: Neo4j config files
  • data: Neo4j databases
  • logs: Logs generated by Neo4j
  • plugins: Plugins used by Neo4j
• docker-compose.yml: Docker config file
• marinewatch-cli: CLI tool used to manage the app

How it works

Important:

• You need to have docker running

Run Batch Spark job with specified accuracy

./marinewatch-cli -b 40

Run Streaming Spark job listening on specified address

./marinewatch-cli -c

Create Neo4j database

./marinewatch-cli -u dbname
# create new database named dbname
# import new data inside, start the database
# create an index on (latitude,longitude)

Start existing Neo4j database

./marinewatch-cli -d dbname

Start web API

./marinewatch-cli -s
# You can see the result from this endpoint for example
# http://localhost:4567/route?from=39.425,6.825&to=6.225,103.050

Batch processing + New database + Web Server in a single command

./marinewatch-cli -s -b 40 -u dbname

Full CLI Documentation

$ ./marinewatch-cli -h
Usage: marinewatch-cli [-h] [-b <int>] [-c <string>] [-u <string>] [-d <string>] [-s] [-t]

  -h  Help. Display this message and quit.
  -b  <int>  Run batch process with specified accuracy.
  -c  <string>  Run streaming process listening on specified address.
  -u  <string>  Create new database with name.
  -d  <string>  Start database with name.
  -s  Start web server.
  -t  Start streaming server.

Improvements to do

• ✅ Improve speed
• ✅ Use datasets with better accuracy (1/40)
• ✅ Add Spark Streaming processing
• Do not restrict cost to boats density
• ...

License

This project is licensed under the MIT License. See the LICENSE file for details.