Benchmark Results
Predictable Low Latency at High Concurrency
4.96ms
Avg Latency (Point Selects)
25,814
QPS (Point Selects)
~14ms
Avg Latency (Range Scans)
sysbench benchmark, 10M row dataset. Full methodology →
The Old Way
Is Data Platform Complexity Holding Your Real-Time Apps Back?
Serving real-time analytics means stitching together a stream processor and a serving database — a brittle pipeline with consistency issues, high latency, and double the operational overhead.
The RisingWave Way
One System. Always Fresh. Always Consistent.
RisingWave continuously processes data streams using incrementally maintained Materialized Views. Results are stored in its own distributed storage, ready for immediate, low-latency querying via standard SQL.
Architecture
A Smarter Way to Serve Real-Time Data
Unified Processing & Storage
An integrated stream processor and storage layer. Eliminate the Flink+DB stack and cut operational complexity in half.
Full SQL, Not Just Key-Value
Complex aggregations, JOINs, and range scans directly on your freshest data — via standard PostgreSQL wire protocol.
Strong Consistency, Guaranteed
Reads are always strongly consistent. Serve financial-grade, accurate data — the most recent results, every time.
100K QPS, Single-Digit ms Latency
Storage engine optimized for fast data retrieval at high concurrency. Point queries under 5ms, range scans under 15ms.
Use Cases
From Raw Streams to Served APIs in One Step
Real-Time Feature Serving
Compute and serve up-to-the-millisecond features for ML models — like 'user activity in the last 5 minutes' — with a single system.
Live Dashboards & APIs
Power internal BI dashboards or user-facing APIs with data that is always fresh, consistent, and served with low latency.
Microservice State Sharing
Enable services to query a consistent, real-time view of business state without complex service-to-service calls.
Comparison
The Right Tool for the Real-Time Serving Job
vs. Flink + ClickHouse
Simpler Than the Real-Time Stack
Eliminate the complexity of a two-system pipeline. One system for processing and serving — less ops, fresher data.
vs. Redis
More Powerful Than a Cache
Beyond simple key-value gets. Full SQL — aggregations, JOINs, range scans — on automatically freshened data, single-digit ms latency.
vs. PostgreSQL
Built for Serving, Not Just Storage
PostgreSQL familiarity with a distributed, scalable engine built for native stream processing and low-latency serving.
