Distributed Tracing Platform

Part of the worlds-biggest-software-project initiative.

An OpenTelemetry-native distributed tracing platform that gives engineers end-to-end visibility into requests as they cross service boundaries — without enterprise pricing or vendor lock-in.

Modern systems decompose into dozens of services per request, and traditional logs and metrics cannot reconstruct the full call graph when latency spikes or errors cascade. This project builds a tracing platform for engineering teams that need spans, traces, service dependency maps, and latency analysis at scale, with cost-efficient retention and AI-assisted debugging.

Why Distributed Tracing Platform?

Observability spend is growing fast. The market is projected to reach roughly USD 4.35 billion by 2026 with a 16.5% CAGR through 2035, driven by microservices and cloud-native adoption — yet teams remain trapped between expensive SaaS and operationally heavy self-hosted stacks.
High-cardinality querying is gated behind proprietary SaaS. Honeycomb's wide-events model is the gold standard for ad-hoc debugging, but is cloud-only with no self-hosted option and higher per-GB cost than self-hosted alternatives at scale.
Open-source incumbents have UX and feature gaps. Jaeger's web UI is functional but less polished than commercial tools and lacks built-in anomaly detection or SLO management. Grafana Tempo requires a separate Grafana instance and has no native alerting on trace data.
Cloud-native tracing locks teams into a single vendor. AWS X-Ray ties instrumentation to AWS and constrains cardinality on indexed annotations, making cross-cloud or migration paths painful.
AI-augmented tracing is largely proprietary. Automated root-cause hints (Dynatrace Davis, X-Ray Insights, Honeycomb BubbleUp) sit inside closed platforms; an open AI-native equivalent does not yet exist.

Key Features

OpenTelemetry-Native Ingest

OpenTelemetry SDK compatibility for traces, with logs and metrics correlation via shared trace and span IDs
Native acceptance of OpenTelemetry, Jaeger, and Zipkin trace formats
OpenTelemetry Collector as the standard ingest path
Tail-based sampling with configurable rules (always-sample errors, slow traces)

Storage and Retention

Object-storage backend (S3, GCS, Azure Blob) for cost-efficient retention at scale
Multi-tenant data isolation for SaaS deployments
Tiered retention strategies to balance fidelity against cost
Pluggable backend support patterns proven by Jaeger (Cassandra, Elasticsearch, ClickHouse)

Query and Visualisation

Waterfall trace viewer with span detail and expandable metadata
Service dependency / call-graph map derived from span data
TraceQL-compatible query language for programmatic trace exploration
Trace filtering by user, error type, response time, and custom attributes
Comparative trace diffing between time windows or deployments

Alerting and SLOs

Threshold-based alerting on p50/p95/p99 latency and error rates per service
SLO definition and burn-rate dashboards surfaced from trace data
Anomaly detection highlighting attribute combinations correlated with degraded performance
Integration hooks for log backends (Loki, Elasticsearch, CloudWatch) via trace ID linking

AI-Augmented Debugging

Automated root-cause suggestion from anomalous trace patterns
Natural-language trace querying (e.g. "show me all traces where the payment service was slow last Tuesday")
Predictive alerting before SLO burn rates reach critical thresholds
Span attribute enrichment, inferring missing context from correlated data

AI-Native Advantage

Incumbent AI features — BubbleUp, X-Ray Insights, Dynatrace Davis — are powerful but locked inside proprietary platforms. An open AI-native tracing platform can deliver natural-language trace queries, automated root-cause suggestions over anomalous span patterns, and predictive SLO burn-rate alerting on top of OpenTelemetry-standard data. Combined with span-attribute enrichment from correlated signals, AI shifts tracing from manual waterfall inspection to guided debugging — without the trade-secret algorithms or per-seat pricing of commercial APM.

Tech Stack & Deployment

Instrumentation: OpenTelemetry SDKs (the de-facto vendor-neutral standard) and OpenTelemetry Collector pipelines
Storage: Object-storage-first design (S3, GCS, Azure Blob) following the operational model proven by Grafana Tempo, with optional pluggable backends (Cassandra, Elasticsearch, ClickHouse) as demonstrated by Jaeger v2
Deployment modes: Self-hosted (Kubernetes Operator for automated rollout) and managed/cloud variants for teams that prefer not to run infrastructure
Standards alignment: OpenTelemetry for ingest; TraceQL-compatible query semantics; trace/span ID propagation for cross-signal correlation

Market Context

The observability tooling market is projected at roughly USD 4.35 billion by 2026, growing at 16.5% CAGR through 2035 (research.md). Buyers span platform engineering teams, SREs, and engineering leaders responsible for reliability of microservice estates. Pricing today ranges from free open source (Jaeger under Apache 2.0; Tempo under AGPLv3) through usage-priced cloud (AWS X-Ray) to enterprise SaaS (Honeycomb, Dynatrace), with cost-at-scale a recurring pain point across the segment.

Project Status

This project is in the research and specification phase.
Contributions, feedback, and domain expertise are welcome.

Contributing

We welcome contributions from developers, domain experts, and potential users. See CONTRIBUTING.md for guidelines.

Important: All contributions must be your own original work or clearly attributed open-source material with a compatible licence. Copyright infringement and licence violations will not be tolerated and will result in immediate removal of the offending contribution. If you are unsure whether a piece of code, text, or other material is safe to contribute, open an issue and ask before submitting.

Licence

Licence to be determined. See discussion for context.