5.52M events/sec
Lock-free MPMC ring buffer with Highway SIMD batch copy. Zero allocation on the hot path.
Features
Ingestion Engine
Section titled “Ingestion Engine”Feed Handlers
Built-in parsers for FIX 4.4, ITCH, Binance WebSocket, and Kafka. Plug in custom feeds via C++ API.
WAL
Write-ahead log for crash recovery. Configurable sync policy (every write, periodic, or async).
Query Engine
Section titled “Query Engine”LLVM JIT
Queries compiled to native code at runtime. Vectorized execution with SIMD aggregation.
ASOF JOIN
Point-in-time correct joins for financial and sensor data. Nanosecond precision.
Window JOIN
Join within a time window (±N ms). Essential for sensor fusion and event correlation.
Built-in Analytics
EMA, VWAP, xbar, mavg, percentile — temporal analytics functions built into the engine.
Standard SQL
Full SQL support: SELECT, JOIN, GROUP BY, window functions, subqueries, CTEs.
Storage
Section titled “Storage”| Feature | Description |
|---|---|
| In-Memory Column Store | Columnar layout optimized for SIMD scan and aggregation |
| Arena Allocator | Custom memory management — no malloc on hot path |
| Partition Manager | Symbol-based partitioning for data locality |
| Parquet HDB | Historical database on S3/NFS. Query hot + cold data transparently |
| Compression | LZ4 for WAL, Parquet columnar compression for HDB |
Client APIs
Section titled “Client APIs”| API | Latency | Use Case |
|---|---|---|
| Python (zero-copy) | 522ns | Quant research, ML pipelines, Jupyter notebooks |
| C++ Pipeline | Native | Embedded, low-latency applications |
| HTTP REST | ~1ms | Web apps, microservices, Grafana |
| Arrow Flight | ~100μs | Batch data transfer, distributed training |
| SQL (CLI) | Interactive | Ad-hoc queries, debugging |
Security
Section titled “Security”- TLS 1.3 with optional mTLS
- RBAC with 5 built-in roles (admin, writer, reader, analyst, monitor)
- JWT / OIDC authentication (Okta, Auth0, Azure AD, Google)
- API key management with rotation
- Audit logging (all auth events, queries, admin actions)
- Rate limiting per user/role
- Multi-tenancy with namespace isolation
- SOC2 & MiFID II compliance ready
See the Security Operations Guide for details.
Clustering
Section titled “Clustering”- Multi-node with automatic sharding by symbol
- Consistent hashing for partition placement
- Health monitoring and automatic failover
- Rolling upgrades with zero downtime
- Cross-node query routing and aggregation
See Multi-Node Cluster for setup.
Deployment
Section titled “Deployment”- Docker (single command start)
- Kubernetes / Helm
- Bare metal with NUMA/CPU pinning optimization
- ARM Graviton support (aarch64)
- Parquet HDB on S3, GCS, or local NFS
See Production Deployment for details.