Event‑Driven Messaging Patterns at Wix: Consumption, Projection, End‑to‑End Streaming, In‑Memory KV Stores, Scheduling, Transactions, and Aggregation

Over the past year the author worked on Wix's event‑driven messaging infrastructure built on Kafka, serving more than 1,400 microservices and providing a scalable backbone for growing traffic and storage needs.

1. Consumption and Projection – To alleviate bottlenecks in popular services like MetaSite, Wix streams all site‑metadata objects to a Kafka topic, creates a write‑only service that projects only the installed‑app‑context into a dedicated database, and a read‑only service that serves queries from this materialized view.

2. End‑to‑End Event‑Driven Flow – By combining Kafka with WebSocket managers, Wix replaces request‑reply polling with a fully distributed, fault‑tolerant pipeline for importing contacts, delivering real‑time status updates to browsers without maintaining server‑side state.

3. In‑Memory KV Store – Compressed Kafka topics act as persistent, zero‑latency key/value stores, enabling services such as Business Manager, Bookings, and Events to share configuration data (e.g., country lists, time‑zones) instantly across data centers.

4. Schedule‑and‑Forget – Recurring jobs like subscription renewals are orchestrated by publishing requests to Kafka instead of repeatedly calling REST endpoints, leveraging Kafka’s ordered processing and built‑in retry mechanisms to ensure eventual success without heavy DB locking.

5. Transactions in Events – Exactly‑once semantics are achieved by wrapping checkout processing in Kafka producer transactions, guaranteeing that downstream services (inventory, delivery, invoicing) see each event only after the transaction commits.

6. Event Aggregation – Using compacted topics as an atomic KV store, Wix tracks the completion of many parallel import jobs, updating a job‑completed counter in a deterministic order and notifying users via WebSocket once all jobs finish.

The overall design demonstrates how Kafka enables ordered processing, reduces retry frequency, improves fault tolerance, and decouples microservices while maintaining strong consistency guarantees.