Write complex SQL queries.
Add data from a message bus like Kafka.
Your SQL views stay incrementally updated within milliseconds.
At millions of updates per second.
No manually glueing streams. No writing imperative code.
Declare what you want in SQL. Materialize keeps it up to date. Within milliseconds.
No "inspired by SQL" half-broken query languages. Materialize supports ANSI Standard SQL 92. Use a SQL shell or your favorite database adapter.
No page long caveats about eventual consistency and out-of-order execution. Materialize results are always consistent and always up-to-date. Pretend that we reran every query every message.
Materialize is built on top of Timely Dataflow, which is distributed. Materialize also adds active-active cluster replication for fault-tolerance, and autoscaling orchestration.
Scale up: process millions of messages per second on a single-digit number of machines. Scale down: run Materialize on your laptop with a single binary on a single core. Either way, it's likely faster than the competition.
Arjun was an engineer on the SQL team at Cockroach Labs, where he worked on performance and benchmarking. He has a blog.
Frank was at Microsoft Research Silicon Valley where he led the Naiad project and co-invented Differential Privacy. He has a GitHub repository named blog.
Cuong was the third engineer at YouTube, the founding head of Dropbox's first remote office in New York City, and an engineering manager at Cockroach Labs.
Nikhil specializes in distributed systems, SQL databases, and build systems. He was previously an engineer on the core team at Cockroach Labs, where he worked on the replication engine.