THE ARCHITECTURE OF GRAB'S DATA LAKE

Summary from the TLDR Web Dev newsletter

Grab uses a data lake to store and analyze data from its various services. The company carefully selects data storage formats based on data throughput, using different strategies for high and low throughput scenarios. For high-throughput data, it uses Apache Avro with Merge on Read for faster write speeds. For low-throughput data, it uses Parquet with Copy on Write for faster read speeds.