LinkedIn's architecture pioneered many patterns now common in tech. Kafka came from here. So did Voldemort, Espresso, Samza. Here's the professional-network stack.

Member AppFrontend TierPlay frameworkFeed ServiceSearch (Galene)MessagingRecruiter APILearning PlatformEspressoMySQL wrapperKafkaFirehoseVoldemort KVSamzastream proc
LinkedIn: Espresso for OLTP + Kafka for events + Voldemort for KV + Samza streams
Advertisement

Espresso as primary DB

Sharded MySQL wrapper with strong consistency + secondary indexes. Handles member profiles, connections, messaging.

Espresso as primary DB

Sharded MySQL wrapper with strong consistency + secondary indexes. Handles member profiles, connections, messaging.

Advertisement

Kafka as central nervous system

Originally invented at LinkedIn. Every service event → Kafka topic. Downstream: analytics, search, feed, notifications.

Voldemort for KV

LinkedIn's Dynamo-style KV. Used for feature caching, denormalized views. Read-optimized. Rebuilt daily from source of truth.

Samza for streams

Kafka-first streaming (predates Flink). Powers feed personalization, spam detection, real-time analytics.

Galene search

Custom Lucene wrapper. Powers people search + typeahead. Handles 500M+ queries/day with sub-100ms latency.

Espresso + Kafka + Voldemort + Samza + Galene. LinkedIn's stack seeded much of modern data infra.