Cassandra
Distributed wide-column NoSQL — masterless, linearly scalable, designed for massive write volumes across many data centres.
Built at Facebook (2008), open source under Apache. Borrows its storage model from Google BigTable and replication from Amazon Dynamo (consistent hashing, gossip protocol, tunable consistency per query). LSM-tree on disk. Its query language CQL looks like SQL but lacks JOINs — you model your queries up front and denormalise.
Brilliant for extreme write throughput (Netflix, Discord, Apple). Painful for ad-hoc queries or schema changes. ScyllaDB is a C++ reimplementation that speaks the same protocol.