Cloud data lake on GCP

Project Summary

This project creates a data lake on Google Cloud Platform with main focus on building a data warehouse and data pipeline. The data lake is built around the I94 Immigration data provided by the US government. The data warehouse will help US official to analyze immigration patterns to understand what factors drive people to move.

Project Scopes

The scope of this project will be to build a data ware house on Google Cloud Platform that will help answer common business questions as well as powering dashboards. To do that, a conceptual data model and a data pipeline will be defined.

Data sources

I94 Immigration Data: This data comes from the US National Tourism and Trade Office.
I94 Data dictionary: Dictionary accompanies the I94 Immigration Data
World Temperature Data: This dataset came from Kaggle.
U.S. City Demographic Data: This data comes from OpenSoft. You can read more about it here.
Airport Code Table: This is a simple table of airport codes and corresponding cities. It comes from here.

Architecture

Data are uploaded to Google Cloud Storage bucket. GCS will act as the data lake where all raw files are stored. Data will then be loaded to staging tables on BigQuery. The ETL process will take data from those staging tables and create data mart tables. An Airflow instance can be deployed on a Google Compute Engine or locally to orchestrate the pipeline.

Here are the justifications for the technologies used:

Google Cloud Storage: act as the data lake, vertically scalable.
Google Big Query: act as data base engine for data warehousing, data mart and ETL processes. BigQuery is a serverless solution that can easily and effectively process petabytes scale dataset.
Apache Airflow: orchestrate the workflow by issuing command line to load data to BigQuery or SQL queries for ETL process. Airflow does not have to process any data by itself, thus allowing the architecture to scale.

Data Model

The database is designed following a star-schema principal with 1 fact table and 5 dimensions tables.

F_IMMIGRATION_DATA: contains immigration information such as arrival date, departure date, visa type, gender, country of origin, etc.
D_TIME: contains dimensions for date column
D_PORT: contains port_id and port_name
D_AIRPORT: contains airports within a state
D_STATE: contains state_id and state_name
D_COUNTRY: contains country_id and country_name
D_WEATHER: contains average weather for a state
D_CITY_DEMO: contains demographic information for a city

Data pipeline

This project uses Airflow for orchestration.

A DummyOperator start_pipeline kick off the pipeline followed by 4 load operations. Those operations load data from GCS bucket to BigQuery tables. The immigration_data is loaded as parquet files while the others are csv formatted. There are operations to check rows after loading to BigQuery.

Next the pipeline loads 3 master data object from the I94 Data dictionary. Then the F_IMMIGRATION_DATA table is created and check to make sure that there is no duplicates. Other dimension tables are also created and the pipelines finishes.

Scenarios

Data increase by 100x

Currently infrastructure can easily supports 100x increase in data size. GCS and BigQuery can handle petabytes scale data. Airflow is not a bottle neck since it only issue commands to other services.

Pipelines would be run on 7am daily. how to update dashboard? would it still work?

Schedule dag to be run daily at 7 AM. Setup dag retry, email/slack notification on failures.

Make it available to 100+ people

BigQuery is auto-scaling so if 100+ people need to access, it can handle that easily. If more people or services need access to the database, we can add steps to write to a NoSQL database like Data Store or Cassandra, or write to a SQL one that supports horizontal scaling like BigTable.

Project Instructions

GCP setup

Follow the following steps:

Create a project on GCP
Enable billing by adding a credit card (you have free credits worth $300)
Navigate to IAM and create a service account
Grant the account project owner. It is convenient for this project, but not recommended for production system. You should keep your key somewhere safe.

Create a bucket on your project and upload the data with the following structure:

gs://cloud-data-lake-gcp/airports/:
gs://cloud-data-lake-gcp/airports/airport-codes_csv.csv
gs://cloud-data-lake-gcp/airports/airport_codes.json

gs://cloud-data-lake-gcp/cities/:
gs://cloud-data-lake-gcp/cities/us-cities-demographics.csv
gs://cloud-data-lake-gcp/cities/us_cities_demo.json

gs://cloud-data-lake-gcp/immigration_data/:
gs://cloud-data-lake-gcp/immigration_data/part-00000-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00001-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00002-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00003-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00004-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00005-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00006-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00007-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00008-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00009-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00010-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00011-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00012-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet
gs://cloud-data-lake-gcp/immigration_data/part-00013-b9542815-7a8d-45fc-9c67-c9c5007ad0d4-c000.snappy.parquet

gs://cloud-data-lake-gcp/master_data/:
gs://cloud-data-lake-gcp/master_data/
gs://cloud-data-lake-gcp/master_data/I94ADDR.csv
gs://cloud-data-lake-gcp/master_data/I94CIT_I94RES.csv
gs://cloud-data-lake-gcp/master_data/I94PORT.csv

gs://cloud-data-lake-gcp/weather/:
gs://cloud-data-lake-gcp/weather/GlobalLandTemperaturesByCity.csv
gs://cloud-data-lake-gcp/weather/temperature_by_city.json

You can copy the data to your own bucket by running the following:

gsutil cp -r gs://cloud-data-lake-gcp/ gs://{your_bucket_name}

Local setup

Clone the project, create environment, install required packages by running the following:

git clone https://github.com/tuanchris/cloud-data-lake
cd cloud-data-lake

Install docker if it's not already installed. You can find the resources to do that here.

Install the Astronomer CLI following the instructions here.

Run the following commands to bring up the Airflow instance:

astro d start

You can look at the logs by running make logs if you need to debug something. You can access and manage the pipeline by typing the following address to a browser:

localhost:8080/admin/

If everything is setup correctly, you will see the following screen:

Navigate to Admin -> Connections and paste in the credentials for the following two connections: bigquery_default and google_cloud_default

Navigate to the main dag on path dags\cloud-data-lake-pipeline.py and change the following parameters with your own setup:

project_id = 'cloud-data-lake'
staging_dataset = 'IMMIGRATION_DWH_STAGING'
dwh_dataset = 'IMMIGRATION_DWH'
gs_bucket = 'cloud-data-lake-gcp'

You can then trigger the dag and the pipeline will run.

The data warehouse

The final data warehouse looks like this: