Platform · 2008–now · Cloud Scale
Apache Cassandra
Wide-column store built for Netflix-scale write throughput. Resume-driven clusters optional.
Why Apache Cassandra stuck
Cassandra stuck at organizations that genuinely needed multi-datacenter write scaling and could afford Cassandra admins. It failed as a default because operability and repair stories terrify teams with gigabyte problems. The graveyard is full of Cassandra clusters serving Postgres workloads.
Compare with
Related
Practice · 2009–2019
NoSQL for Everything
Schema-optional databases sold as a lifestyle. Right tool for some jobs; default for none.
$ Startups ditched Postgres for document stores "because scale," then rebuilt relational integrity in application code — and paid twice when joins came back as a product requirement.
Platform · 2006–2018
Hadoop Everywhere
MapReduce as lifestyle. The elephant in rooms that needed a spreadsheet.
$ Enterprises stood up Hadoop clusters for gigabytes of data that fit on one Postgres instance. Hadoop admins, ZooKeeper nightmares, and ETL rewrites burned millions before Spark and cloud warehouses retired the elephant.
Platform · 2011–now
Apache Kafka
Durable log as nervous system. Event-driven architecture with real operational teeth.