Documentation Index
Fetch the complete documentation index at: https://mintlify.com/timeplus-io/proton/llms.txt
Use this file to discover all available pages before exploring further.
Overview
Timeplus Proton is the fastest SQL pipeline engine in a single C++ binary, for stream processing, analytics, observability and AI. It’s a simple, fast and efficient alternative to ksqlDB and Apache Flink, powered by the ClickHouse engine.Timeplus Proton is open source under Apache License 2.0, with no JVM, no ZooKeeper, and zero dependencies.
Key Features
Blazing Fast Performance
Written in C++ with SIMD optimizations. Delivers 90 million events per second with 4ms end-to-end latency on an M2 Max MacBook Pro.
Lightweight & Efficient
Single binary under 500MB. No JVM required. Runs on AWS t2.nano (1 vCPU, 0.5 GiB memory).
SQL for Everything
Native sources/sinks for Kafka, ClickHouse, MySQL, Postgres, MongoDB, S3/Iceberg, OpenSearch. Streaming ingestion, multi-stream JOINs, materialized views, and UDFs in Python/JavaScript.
Powered by ClickHouse
Extends ClickHouse with stream processing capabilities. Thousands of SQL functions available. Query billions of rows in milliseconds.
Why Choose Timeplus Proton
Alternative to Apache Flink and ksqlDB
Timeplus Proton provides powerful stream processing functionalities including:- Streaming ETL pipelines
- Tumble/hop/session windows
- Watermarks and late event handling
- Incremental materialized views
- CDC and data revision processing
- Queryable analytical and row-based materialized views
Performance Advantages
- Speed
- Resource Efficiency
- Kafka Integration
Timeplus Proton is written in C++ with optimized performance through SIMD.Benchmark results (Apple MacBook Pro M2 Max):
- 90 million events per second throughput
- 4 millisecond end-to-end latency
- 1 million unique keys for high cardinality aggregation
Common Use Cases
Timeplus Proton empowers you to build a wide range of real-time applications and data pipelines:Real-time Analytics ETL/Pipeline
Efficiently ingest live data from sources like Kafka, perform in-pipeline transformations (filtering, enrichment, masking), and route it to downstream systems including:- Data warehouses like ClickHouse
- Other Kafka topics
- Analytical stores
- Time-series databases
Real-time Telemetry Pipeline and Alerting
Process and route logs, metrics, and traces with:- In-pipeline noise reduction
- Real-time alerts before forwarding
- Integration with Splunk, Elastic, or S3
- Custom transformation rules
Real-time Feature Pipeline for AI/ML
Compute real-time features using:- Low-latency, high-throughput streaming SQL
- Materialized views with backfill support
- Advanced windowing over live data
- Direct integration with ML platforms
Architecture Overview
Data Storage
Proton includes two complementary storage engines:- Streaming Store: Optimized for append-only, high-throughput writes and streaming queries
- Historical Store: Based on ClickHouse, optimized for OLAP queries on historical data
Query Processing
Timeplus Proton supports both:- Streaming queries: Continuous queries that process data as it arrives
- Historical queries: Traditional batch queries on stored data
- Hybrid queries: Combine streaming and historical data in a single query
Comparison with Other Solutions
| Feature | Timeplus Proton | Apache Flink | ksqlDB |
|---|---|---|---|
| Language | C++ | Java | Java |
| Binary Size | < 500MB | ~300MB+ | ~100MB+ |
| Dependencies | None | JVM, ZooKeeper | JVM, Kafka |
| Deployment | Single binary | Cluster required | Kafka dependent |
| SQL Support | Full ANSI SQL + streaming extensions | Table API + SQL | Limited SQL |
| Performance | 90M EPS | 10-20M EPS | 5-10M EPS |
| Memory | Runs on 0.5GB | Requires 2GB+ | Requires 1GB+ |
Performance numbers are approximate and vary based on workload and hardware configuration.
What’s Next
Quick Start
Get Timeplus Proton running in 5 minutes
Installation
Detailed installation instructions for all platforms
Examples
Explore real-world examples and use cases
SQL Reference
Complete SQL reference documentation