> For the complete documentation index, see [llms.txt](https://docs.pharos.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.pharos.ai/open-telemetry/understanding-otel.md).

# Understanding OpenTelemetry

A product-neutral primer on OpenTelemetry — the three signals, OTLP, semantic conventions, and why backends are interchangeable.

OpenTelemetry (OTel) is the standard your observability platform almost certainly already speaks. Before looking at how Pharos participates in it, it helps to understand the standard itself.

***

## What OpenTelemetry is

OpenTelemetry is a vendor-neutral standard for generating, collecting, and exporting telemetry, governed by the Cloud Native Computing Foundation (CNCF). It defines a common data model and a common wire protocol, so instrumentation and backends don't need to agree on anything except OTel itself. A source emits data in the OTel format; any OTel-compatible backend can receive it, regardless of who built the instrumentation or who built the backend.

***

## The three signals

OpenTelemetry defines three telemetry signals:

| Signal      | What it captures                                                                                                                    |
| ----------- | ----------------------------------------------------------------------------------------------------------------------------------- |
| **Traces**  | A tree of **spans** — timed units of work — showing how a single transaction moved through your systems and how long each step took |
| **Metrics** | Numeric measurements over time — counts, durations, gauges — for volume, latency, and resource consumption                          |
| **Logs**    | Discrete, timestamped records — events, errors, audit facts — correlated to a trace when relevant                                   |

A single underlying event can produce more than one signal at once: a span for its execution, a metric for its volume or latency contribution, and a log if it represents a notable fact — an error, an audit event — worth recording on its own.

***

## OTLP

OTLP (the OpenTelemetry Protocol) is the wire format signals travel over. It's a defined gRPC or HTTP protocol, so any exporter that speaks OTLP can send data to any backend that receives OTLP — there's no custom integration to build for each pairing. A backend that accepts OTLP can ingest traces, metrics, and logs from any OTel-instrumented source.

***

## Semantic conventions

OpenTelemetry standardizes not just the transport but the vocabulary. **Semantic conventions** are shared attribute names — `http.status_code`, `db.system`, and the like — that instrumentation across languages and vendors agrees to use. When telemetry follows semantic conventions consistently, it stays portable: a backend's dashboards, alerts, and correlations work the same way regardless of which tool produced the underlying data.

***

## Vendor-neutrality

Because the protocol and data model are standardized, backends become interchangeable **OTLP routes** rather than proprietary integrations. You can point the same OTel pipeline at Honeycomb, Splunk, Datadog, or any other OTLP-compatible platform — and change your mind later — without re-instrumenting anything upstream. This is the property that makes OpenTelemetry attractive to teams who don't want to be locked into a single observability vendor.

***

## Related

<table data-view="cards"><thead><tr><th></th><th></th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td><strong>🤝 OpenTelemetry &#x26; Pharos</strong></td><td>How Triton and Hermes participate in the standard</td><td><a href="https://docs.pharos.ai/open-telemetry/understanding-otel/otel-and-pharos.md">OpenTelemetry &amp; Pharos</a></td></tr></tbody></table>
