Noria serves millions of requests per second on a single server by integrating an in-memory cache and a SQL database in a single dataflow computation that incrementally maintains materialized views for application queries.

A new dataflow model, partially-stateful dataflow, helps Noria achieve this performance with bounded memory and computation cost. Unlike prior streaming dataflow systems, which store and update full state in each stateful operator (\eg joins, aggregations, and materialized views), Noria stores and maintains only a subset of it.

Partial state also allows Noria to install new queries in its live dataflow in milliseconds, so applications can evolve and pose new queries without any backend downtime.

This is joint research project with the PDOS group at MIT.

Publications

  • Noria: dynamic, partially-stateful data-flow for high-performance web applications
    Gjengset, Jon, Schwarzkopf, Malte, Behrens, Jonathan, Araújo, Lara Timbó, Ek, Martin, Kohler, Eddie, Kaashoek, M. Frans Kaashoek, and Morris, Robert
    In Proceedings of the 13th USENIX Symposium on Operating Systems Design and Implementation (OSDI)
    2018.
    [PDF]