Introducing ZeptoDB

The time-series database market has a gap. kdb+ delivers microsecond performance but costs $100K+/year and requires learning a proprietary language. ClickHouse and InfluxDB are affordable but operate at millisecond latency — 100-1000x slower for real-time workloads. None offer zero-copy Python integration for ML pipelines.

ZeptoDB closes this gap.

---

The Architecture

ZeptoDB is a C++20 in-memory columnar database built from scratch for microsecond-latency time-series workloads. The core pipeline:

Feed Handler → MPMC Ring Buffer → Column Store → SIMD/JIT Query Engine
                    ↓
                   WAL (crash recovery)
                    ↓
               Parquet HDB (S3/NFS)

Every layer is designed around a single principle: zero-copy, zero-allocation on the hot path.

• Ingestion: Lock-free MPMC ring buffer with Highway SIMD batch copy — 5.52M events/sec
• Storage: Arena allocator with columnar layout, no malloc during ingestion or query
• Execution: LLVM JIT compiles SQL to native code, Highway SIMD for vectorized aggregation
• Python: pybind11 zero-copy — NumPy arrays reference engine memory directly (522ns)

---

How It Compares

	kdb+	ClickHouse	InfluxDB	ZeptoDB
Latency	μs	ms	ms	μs
Ingestion	~5M/sec	100K/sec	50K/sec	5.52M/sec
SQL	✗ (q lang)	✓	InfluxQL	✓ Standard SQL
Python	IPC	—	—	522ns zero-copy
ASOF JOIN	✓	✗	✗	✓
License	$100K+/yr	OSS	OSS	OSS (Apache 2.0)

---

Who Is It For?

Quant Researchers

ASOF JOIN, VWAP, EMA, xbar — the temporal operations you need, in standard SQL. Python zero-copy for notebooks.

Trading Desks

5.52M ticks/sec ingestion with built-in FIX 4.4, ITCH, and exchange WebSocket parsers.

IoT / Robotics

10KHz sensor ingestion, Window JOIN for sensor fusion, Parquet HDB for historical replay.

---

What’s Next

ZeptoDB is open source under Apache 2.0. We’re actively developing:

• MQTT / OPC-UA connectors for industrial IoT
• JDBC/ODBC drivers for BI tool integration
• Managed cloud offering

Try it: Quick Start →