Distributed tracing uses a small set of related concepts to describe a request and its lifecycle. Understanding these terms makes trace search, topology, sampling, and retention behavior easier to interpret.
Concepts in this section
At a glance
| Term | Meaning |
|---|
| Trace | The end-to-end record of a request or transaction. |
| Span | A timed unit of work within a trace. |
| Root Span | A span with no parent within its trace. It normally represents the entry point; missing spans can leave the displayed trace incomplete. |
| Child Span | A span whose parent span ID identifies another span in the same trace. |
| Trace ID | Identifier shared by all spans in the same trace. |
| Span ID | Identifier for one span. |
| Parent Span ID | Identifier connecting a child span to its caller. |
| Application | A logical grouping of services used to organize trace investigation in OpsRamp. |
| Service | Logical application component that performs work. |
| Resource | Entity that produced telemetry, such as a process, host, pod, or container. |
| Operation | Named activity represented by a span. |
| Span Kind | The role of the operation, such as server, client, producer, consumer, or internal application work. It is distinct from transaction type and category. |
| Attribute | Searchable key-value metadata. |
| Event | Timestamped occurrence recorded during a span. |
| Status | Outcome reported for a span, commonly unset, successful, or error. |
| Error Indicator | A marker used to highlight a span classified as failed. Interpret it alongside span status and protocol-specific attributes such as HTTP response status. |
| Trace Context | The trace identity, active span identity, and tracing options carried between related operations. |
| Propagation | Injecting and extracting trace context across application, request, or messaging boundaries. |
| Sampling | Selecting which traces are recorded or forwarded. Traces dropped by sampling cannot be recovered by extending retention. |
| Ingestion | Acceptance of exported spans for the intended OpsRamp client. |
| Retention | The period for which accepted trace data remains available. |
OpsRamp Classifications
Use these classifications to narrow service, operation, and span views. Their values depend on the telemetry supplied by the application and the collection configuration.
| Term | Meaning |
|---|
| Transaction Type | A Web or Non-Web classification derived from the recorded operation's attributes. It is not the same as Span Kind. |
| Transaction Category | The technology or protocol associated with the operation, such as HTTP, Database, RPC Systems, or Messaging queues. |
| Transaction Subcategory | A more specific value within the category. For HTTP, it can be a method such as GET or POST; for Database, it can identify the database system, such as PostgreSQL; for Messaging queues, it can identify the messaging system. It is not universally an action such as SELECT or INSERT. |
Interpret metrics within the same application, service, operation, filters, and time range. Trace-derived metrics describe observed traffic and may not represent requests omitted by the collection or sampling configuration.
| Term | Meaning |
|---|
| Trace Duration | Elapsed time covered by a trace. It is not the sum of all span durations because spans can overlap. |
| Latency / Average Latency | Duration of the selected operation; average latency summarizes durations across the observed operations. Check the unit displayed by the view. |
| Max Latency | The longest observed duration in the selected scope and time range. |
| Median / p50 | The latency at or below which approximately 50% of observed durations fall. |
| p95 / p99 | Latency percentiles at or below which approximately 95% or 99% of observed durations fall. They help identify slow-tail behavior. |
| Throughput / RPM | The rate of observed requests or operations. RPM means requests per minute. |
| Max Throughput | The highest throughput observed across the displayed time intervals, not a total request count. |
| Total Count | The operation count represented by the metrics for the selected scope and time range. It is not necessarily a count of unique traces. |
| Error Rate (%) | The percentage of observed requests or operations classified as errors under the view's error criteria. Check HTTP error-code settings and span status when interpreting it. |
| SL Success (%) | The percentage of eligible HTTP requests meeting the view's success criterion. Compare it with Error Rate only when both measures use the same request population and error criteria. |
| SL Latency (%) | The percentage of eligible operations with latency at or below the configured latency threshold. It is a threshold-based measure, not a latency percentile. |
| Apdex | A score from 0 to 1 that summarizes response-time satisfaction using configured satisfied and tolerated thresholds. Higher scores indicate a larger share of requests meeting those targets; it is not a direct measurement of user sentiment. |
Investigation Views
| Term | Meaning |
|---|
| Trace Explorer | A view for finding traces or spans using a time range, filters, and queries. |
| Timeline / Waterfall | A view of span start times, durations, and parent-child relationships within a trace. It helps identify nesting, concurrency, and delays. |
| Service Maps | A dependency view built from observed service relationships. A trace-specific map shows the dependencies represented in that trace. |
| Worst Spans | A filter for spans marked as errors or exceeding the selected duration threshold. |
| Custom Views | Saved trace investigation queries and view settings that can be reopened for repeated analysis. |