Tracing Without Limits
OpenWit lets you store and search every trace with ease, using custom indexes for complex data, Parquet’s columnar compression for efficient cloud storage, and a distributed, decoupled architecture that scales smoothly as data grows.
Outage Investigation in Seconds
- Logs, metrics, and traces are stored together in one system, making cross-signal correlation instant and eliminating tool switching.
- Apache Parquet compresses and organizes telemetry data for fast analytical queries and reduced I/O during investigations.
- Users can decide which datasets to keep hot in memory or on SSDs, ensuring near-instant responses for frequent queries and dashboards.
- New index types can be added as workloads evolve, maintaining flexibility and query efficiency as data grows.
Clean and Consistent Observability Data
- Every incoming batch is validated against predefined schemas, ensuring all telemetry data conforms to structure and type before it’s written to storage.
- Required fields are verified, missing or malformed entries are rejected, and naming conventions are normalized automatically.
- Logs, metrics, and traces follow a consistent schema across services, simplifying queries and analytics across distributed systems.
Deploy Without Changing a Thing
- OpenWit connects directly to your current telemetry stack with no need to rebuild or replace existing infrastructure.
- Ingest data effortlessly from Kafka, OpenTelemetry (gRPC), or HTTP, all processed through the same unified ingestion layer.
- Connect new data sources easily without changing your existing setup.
Handle Peak Traffic with Confidence
- OpenWit’s Rust-based actor system handles massive parallel workloads without blocking or slowdowns.
- Batching and write-ahead logs (WAL) ensure every event is captured safely while maintaining high throughput.
- Automatic deduplication removes repeat events to keep metrics and counts precise, even under heavy load.
Real-Time Dashboards Without Lag
- Built-in caching keeps your most recent and frequently queried data instantly available.
- Hot data lives in memory or local SSDs, while older data stays compressed in Parquet for efficient retrieval.
- Eliminates delays from cold storage, ensuring repeat queries and dashboards load instantly.
Frequently Asked Questions
How do I deploy Openwit for small vs large workloads?
You can run it in a single-node Monolithic Mode for quick starts or small teams, or in a Distributed Mode with specialized nodes for control, ingest, storage, proxy, cache, and search. Everything is configured via one unified YAML so you scale by adding the node types you need.
How do I send data to Openwit and keep it clean?
Openwit ingests from Kafka, native OpenTelemetry gRPC, custom gRPC, and an HTTP API for simple tests. An ingestion gateway handles auth, schema validation, normalization, and batching. The ingest node deduplicates, writes to a durable WAL, records metadata in Postgres, then streams Arrow batches to storage.
Why are queries fast even on lots of data?
Data lands as Parquet in cloud object storage and Openwit builds indexes like bitmaps, bloom filters, zone maps, and Tantivy full-text. Search uses these indexes plus tiered caching (RAM and SSD) to prune reads, then executes with Apache DataFusion and Ballista. The goal is interactive, often sub-second responses depending on data and filters.