deephaven.io

What It Is

Deephaven is a query engine for real-time data, centered around the idea of “Live Dataframes.” Rather than simply micro-batching real-time streams, it performs incremental computation based on table-change deltas and represents query execution plans as DAGs, allowing downstream nodes to keep updating as data changes. Its official site highlights use cases that require computing, displaying, or transforming large volumes of fast-changing data, with typical customers in high-intensity real-time environments such as finance, exchanges, hedge funds, and crypto exchanges.

Core Capabilities

In terms of functionality, Deephaven covers querying, filtering, sorting, grouping, aggregation, formulas, roll-ups, partitioning, time series, as-of joins, range joins, windows, cumulative calculations, and rolling computations. Supported data sources include Kafka, Redpanda, WebSockets, REST/JSON, logs, Iceberg, Parquet, S3, CSV, Arrow/Arrow Flight, SQL, ODBC/JDBC/ADBC, and more, giving it fairly complete hybrid streaming and batch capabilities. Its language ecosystem is also strong: server-side development supports Python, SQL, and Java, while client APIs cover Python, Java, Go, R, C++, and JavaScript, with efficient communication via Barrage/Arrow Flight/gRPC.

Open Source, Self-Hosting, and Pricing

Deephaven follows an open-core model. The Community edition is free and available on GitHub, and the core engine can run locally; the official site provides Docker startup commands. Open-source building blocks include Barrage, jpy, Web Client UI, VS Code Extension, Deephaven Express, and deephaven.ui. The Enterprise edition is priced at 200K - 2M+ USD/year and includes multi-user support, ACLs, clustering, auditing, SSO, scalable UI, dedicated support, and professional services. It is clearly aimed mainly at large teams with substantial budgets.

Pros and Cons

Its strengths are a well-defined real-time computation model that avoids repeated full recomputation; broad coverage of data sources and language APIs; good integration between analysis and application development through Jupyter, VS Code, Excel, React Web UI, deephaven.ui, and more; and extremely rich documentation covering onboarding, architecture, APIs, performance, and troubleshooting. The downsides are that its concepts and system stack are relatively heavy: DAGs, the JVM, Python bridging, and real-time table types all come with a learning curve. Enterprise-grade permissions, clustering, SSO, auditing, and some UI capabilities require the commercial edition, and the Enterprise pricing threshold is high.

Who It’s For and Access from China

Deephaven is suited to teams working on quantitative research, data science, real-time risk control, trading monitoring, operations monitoring, real-time AI data applications, and high-throughput data applications. It is especially relevant for engineering teams that already have data infrastructure such as Kafka, Parquet, Iceberg, and SQL in place. The official site does not clearly state China access or payment methods, so these remain unknown. If access to GitHub, container images, or overseas services is unstable, users may need to prepare alternative networking options. Comparable products include Apache Flink, Spark Structured Streaming, Materialize, RisingWave, ksqlDB, as well as more application-layer tools such as Streamlit and Dash.