> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aiaxoniq.com/llms.txt
> Use this file to discover all available pages before exploring further.
# APM
> How the APM section turns your spans into projects, request rates, error rates and response times — and which of its views are served today.
APM shows how your services answer requests. Every number in it is computed
from the spans your services send — there is no separate APM agent and no
separate APM data stream.
**Before you start.** You need traces arriving from at least one service —
see [Send data with OpenTelemetry](/send-data/otel/collector). Each service
must set `service.name`, because that is how APM decides which project a span
belongs to.
## Where to start
The APM section opens on three views. Everything else is reached by choosing a
project first.
| View | What it shows |
| :-------------- | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Projects** | Every project on your account and every service reporting spans, with language, response time, requests, failure rate and status. This is the default view. |
| **Trace** | The trace explorer across your whole account. It is the same explorer as the [Traces](/guides/traces/overview) section. |
| **Service map** | How one project's services call each other. It needs a project, so with none chosen it sends you to the project list. See [Service map](/guides/services/service-map). |
## What a project is
A project is a named record that **claims services by name**:
* A project claims the service whose `service.name` equals the project's own
name, with no further setup. A project called `checkout` shows the spans of a
service reporting as `checkout`.
* You can add more services to it under **Manage → Settings → Services**, for a
project named after a business capability that spans several services — for
example `checkout` covering `cart`, `catalogue` and `user`.
A service that sends spans but that no project claims still appears in the
**Projects** list, marked **unregistered**. Clicking it opens its traces, but it
has no project dashboard until a project claims it.
To create a project, choose **Add APM project** on the **Projects** view. The
dialog offers four types — APM, Infrastructure, Logs and Kubernetes — and, for
APM, a language. The type does **not** change which telemetry the project
covers; it is a label and part of the project's identity. Creating the same
name, type and language twice returns the existing project rather than a second
one.
**A service belongs to one project.** Adding a service that another project
already claims is refused with **Already mapped**, naming the project that
holds it. Remove it there first, then add it here.
## Inside a project
Choosing a project opens its dashboard, with a side navigation grouped into
five parts:
| Group | Views |
| :-------------- | :------------------------------------------------------------------------------------------------------------- |
| **Monitor** | Dashboard, Transactions, Errors, HTTP Failures |
| **Investigate** | Session Traces, Service Map, Compare |
| **Catalog** | Services, Databases, External requests |
| **Runtime** | Hosts, Runtime metrics — plus JVM metrics and JVM threads for a Java project, and Go routines for a Go project |
| **Manage** | Deployments, Versions, Reports, Alerts, Settings |
Every view uses the time range in the toolbar. Changing it reloads every panel
together, so no two panels on a page describe different windows.
### The project dashboard
The dashboard answers "is anything wrong" first and "where" second:
* **Golden signals** across the top — services, requests, span error rate and
average response time — each compared with the window immediately before the
current one.
* **Requests over time** and **Errors over time**.
* **Response time**, switchable between the average per service, the time spent
in each layer, and p50/p95/p99 percentiles for the whole project.
* **Status codes** (HTTP responses only) and a **Response time distribution**.
* **Slowest services**, **Error rate by service**, **Apdex** and **Slowest
transactions**.
### How the numbers are counted
| Figure | Counted from |
| :------------------ | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Requests** | Server spans — the inbound side of a request at every hop. A request that makes a database call and calls two services is still one request. |
| **Response time** | The same server spans, as an average and as p95/p99. |
| **Span error rate** | Spans whose OpenTelemetry status is `ERROR`, as a share of **all** spans, not only requests. |
| **Transactions** | Spans that carry both an HTTP method and `http.route`, grouped by service, route and method. |
| **Errors** | Failing spans grouped by their normalised status message, so the same fault with a different id, address or number in its message is one group. |
| **Apdex** | Computed by the server from the latency distribution against the project's **Apdex target (T)**, which you set in **Manage → Settings**. A new APM project starts with 0.5. |
**Requests are not spans.** An earlier way of counting treated every span as a
request, which inflates throughput by each service's fan-out. APM counts
server spans only, so its request figures are lower than the span count in
the trace explorer, and that is expected.
## How APM relates to traces
APM and the trace explorer read the same spans.
* **Session Traces** inside a project lists that project's traces. A trace the
project took part in arrives **whole**, across every service it crossed —
the project decides which traces are listed, not which spans of a trace you
see.
* **Trace** on the account-level strip lists every trace on the account.
* Opening a trace from an error, an HTTP failure or a slow database operation
shows the same waterfall. See [Exploring traces](/guides/traces/overview)
for reading it.
## Views that are not served yet
Some views in the project navigation are built but wait on an API route that
does not exist yet. **External requests**, **JVM metrics**, **JVM threads** and
**Go routines** are among them. These views say so — "This needs …, which the
API does not serve yet" — rather than drawing an empty chart that would look
like an idle service.
## Troubleshooting
A project with no telemetry in the selected window shows **\ is
not reporting yet**. Check that your service sets `service.name` to exactly
the project's name — for an OpenTelemetry SDK, `OTEL_SERVICE_NAME` — or add
the name your service reports under **Manage → Settings → Services**. Then
widen the time range: a service that was quiet in the window has nothing to
show in it. The page fills in once the first span arrives.
**Nothing reporting, and no projects registered** means no service has sent
a span and no project exists yet. A row appears as soon as a service sends a
span or a project is created. See
[Send data with OpenTelemetry](/send-data/otel/collector).
It is sending spans, but no project claims its name. Create a project with
that name, or add the service to an existing project under
**Manage → Settings → Services**.
**Project not found** means the link names a project that does not exist on
this account — it may have been deleted, or belong to another organization.
**projectId is required** means the address holds something other than a
project id, such as a service name. Open the project from the **Projects**
list instead.
Transactions lists spans that carry both an HTTP method and `http.route`. A
span without the route — common when a framework is not auto-instrumented — is still
counted in the dashboard's requests and response time, but it has no route
to be listed under.
## Next
Read the traces behind these numbers.
How your services call each other.
What your users' browsers experience.
Get notified when a number moves.