Datadog automatically generates a set of standard metrics from APM trace data. For measuring latency, it creates distribution metrics. The naming convention for these metrics is trace.. A distribution metric captures statistical information about a set of values, including count, min, max, average, and various percentiles (p50, p75, p99), which collectively represent the latency distribution. Therefore, trace.graphql.execute is the specific distribution metric that represents the latency for all spans named graphql.execute.

A. trace.graphql.execute.apdex is a metric representing the Apdex score, a measure of user satisfaction, not the raw latency distribution.

B. trace.graphql.execute.hits is a count metric that represents the total number of graphql.execute spans, indicating throughput, not latency.

C. trace.graphql.execute.spancount is not a standard suffix for default APM metrics; the correct suffix for a count is .hits.

E. The .distribution suffix is not part of the standard naming convention for these default metrics; the base name trace. itself represents the distribution metric.

1. Datadog Documentation, "APM Metrics". This page lists the standard metrics generated from traces, confirming that trace. is a distribution metric type that tracks the timing of spans.

2. Datadog Documentation, "Metrics Types". This page describes the "Distribution" metric type, explaining that it is used to track the statistical distribution of a set of values, which is ideal for latency measurements.

Watchdog AI is Datadog's automated anomaly detection engine. Its primary benefit is to use machine learning algorithms to continuously analyze metrics, traces, and logs to find unanticipated outliers and abnormal patterns that might indicate a problem. When it detects an anomaly, Watchdog attempts to correlate it with other events, such as code deployments or infrastructure changes, to automatically surface a story that suggests a potential root cause. This significantly reduces the mean time to detection and resolution for complex issues.

A. Watchdog AI detects performance issues; it does not directly modify or improve application code. That is the developer's role, guided by Watchdog's insights.

B. This describes the general data ingestion capability of the Datadog platform, not the specific analytical and anomaly detection function of Watchdog AI.

D. Watchdog AI alerts on degrading metrics but does not directly "improve" them. It provides the necessary context for engineers to take corrective action.

1. Datadog Documentation, "Watchdog." The official documentation states, "Watchdog automatically detects performance problems in your applications, services, and infrastructure, and surfaces a root cause analysis."

2. Datadog Documentation, "Introduction to Watchdog." This page explains that Watchdog "automatically surfaces anomalies and outliers in your telemetry data" and "correlates them to help you find the root cause of issues faster."

The Service Map widget is specifically designed for dashboards to visualize the dependencies and connections between your services. It provides a high-level overview of your application's architecture, showing how services communicate with each other and helping to quickly identify upstream or downstream issues. This widget directly addresses the need to understand data flow and service dependencies.

A. The Service Summary widget displays key performance indicators (requests, errors, latency) for a single, specific service, not the relationships between multiple services.

C. "Service Network Widget" is not an official Datadog widget name. While the Network page shows service connections, this is not a dashboard widget.

D. "Service Architecture Widget" is not an official Datadog widget name. The Service Map is the primary tool for visualizing service architecture.

1. Datadog Documentation, "Dashboards > Widget Library > Service Map": "The Service Map widget displays the connections between services based on their dependencies. Use this widget to get a high-level view of your application architecture."

Datadog Continuous Profiler generates two types of data: raw profiles and profile metrics. Raw profiles, which contain detailed flame graphs and other granular data, are retained for 30 days. Profile metrics are aggregated metrics derived from these profiles (e.g., CPU time by method) and are retained for a much longer period of 15 months, allowing for long-term trend analysis. The question specifically asks for the retention of profile metrics.

A. 7 days is an incorrect retention period for any profiling data.

B. 15 days is the retention period for indexed APM spans, not profile metrics.

C. 1 month (or 30 days) is the retention period for raw profiles, not the aggregated profile metrics.

1. Datadog Documentation, "Continuous Profiler - Data Retention". This page explicitly states: "Profile metrics are stored for 15 months." and "Raw profiles are stored for 30 days." This directly answers the question and differentiates between the two data types.

2. Datadog Documentation, "Continuous Profiler". The overview section often discusses the value of long-term metrics for trend analysis, which is enabled by the 15-month retention period.

Percentile metrics describe the distribution of a value, such as latency. The p100 metric, also referred to as the maximum (max), represents the single highest value recorded in a given time interval. Therefore, p100 latency corresponds to the duration of the single slowest request, capturing the absolute worst-case performance.

B: p99 latency means 99% of requests were faster than this value; it represents the top 1% of slowest requests, not the single slowest.

C: p95 latency means 95% of requests were faster. It is a common measure for tail latency but is not the maximum.

D: p75 latency represents the 75th percentile, meaning 75% of requests were faster than this value.

E: p50 latency is the median; 50% of requests were faster, and 50% were slower.

- Datadog Documentation, "APM Metrics," Section: "Default Metrics." The documentation for trace duration metrics lists aggregations including max, which is equivalent to p100.

- Datadog Documentation, "Guide to graphing with Datadog," Section: "Percentiles." The guide explains the concept of percentiles, where p(N) means N% of values are below that point, making p100 the maximum value.

Datadog Trace Explorer lets you query any span or trace tag by typing @tag:value. No prior facet configuration is required for basic search; facets are only needed to list the tag in the side panel or to build analytics metrics. Therefore you can immediately search with @customertier to locate the relevant spans.

A. Facets are optional for searching; they are needed only for indexing/analytics.

C. The UI supports any custom tag via the @ syntax, not just default tags.

1. Datadog Docs – Trace Explorer, “Search” subsection, sentence 1-3 (https://docs.datadoghq.com/tracing/tracesearchandanalytics/explorer/#search).

2. Datadog Docs – “Facets vs. Search,” same page, ¶2.

The Datadog Continuous Profiler comparison view is designed to identify performance differences within the same service. It allows for comparisons across different time periods, deployments (via the version tag), or any other set of tags (e.g., env:staging vs. env:prod). However, it cannot be used to compare profiles from two different programming languages (e.g., Java vs. Python). Profiling data is specific to a language's runtime and call stack structure, making a direct comparison between different languages meaningless within the tool.

A. Comparing deployments is a primary use case for profile comparison, helping to identify performance regressions introduced by new code.

B. Comparing different time periods is a fundamental feature used to analyze performance changes before and after an event or deployment.

C. Comparing different tag sets (like hosts, regions, or environments) is a key capability for isolating performance issues to specific infrastructure segments.

1. Datadog Documentation, "Continuous Profiler > Compare Profiles". The documentation outlines the comparison functionality, showing options to filter and group by tags like service, version, and env. It implicitly operates within the context of a single service and its language.

2. Datadog Documentation, "Code Hotspots". This page explains that profiles are linked to specific services. Since a service is typically written in a single language, the comparison tool is inherently language-specific.

Datadog provides automatic instrumentation for many popular languages, including Java, Python, and PHP, which requires no code changes. For C++, Datadog provides an OpenTracing-compatible tracer library. This library requires developers to manually add tracing code to their applications to generate spans and traces, meaning it is not automatically instrumented.

A. Java has a powerful auto-instrumentation agent that can be attached to a JVM application without any code modification to trace popular frameworks and libraries.

B. Python's dd-trace-py library can be used with ddtrace-run to automatically patch supported libraries and frameworks to enable tracing without manual code changes.

C. PHP's tracing extension, once enabled in the php.ini file, automatically instruments web requests and supported frameworks, providing tracing with minimal setup.

1. Datadog Documentation, "C++: Tracing Without Limits". Section: "Getting Started". States that developers need to "Add tracing to your code". Retrieved from https://docs.datadoghq.com/tracing/tracecollection/ddlibraries/cpp/.

2. Datadog Documentation, "Tracer Installation & Configuration". This page lists Java, Python, and PHP as having "Automatic instrumentation" capabilities. Retrieved from https://docs.datadoghq.com/tracing/tracecollection/libraryconfig/.

3. Datadog Documentation, "Java: Tracing Without Limits". Section: "Getting Started". Describes attaching the dd-java-agent.jar for automatic instrumentation. Retrieved from https://docs.datadoghq.com/tracing/tracecollection/ddlibraries/java/.

Retention filters are configured exclusively within the Datadog platform's user interface, on the "Retention Filters" page under APM settings. They are a backend mechanism that operates on the entire stream of ingested traces after they have been received by Datadog. These filters determine which traces are indexed and stored for visualization and analytics. They cannot be set at the tracing library or Agent level, as those components are responsible for ingestion controls (sampling), not retention policy.

A. The Tracing Library level is where ingestion sampling rules are configured, not backend retention filters that determine indexing.

B. The Agent level is also for configuring ingestion controls and sampling, not the retention and indexing policies for already-received traces.

C. Since neither the Agent nor the Tracing Library can configure retention filters, this option is incorrect.

1. Datadog Documentation, "Trace Ingestion and Retention". The "Retention Filters" section states: "Retention filters are configured on the Retention Filters page in Datadog. They apply to all traces after they are ingested."

2. Datadog Documentation, "Modifying Traces". This page details processors and filters that can be configured in the Agent, but it distinguishes these from the backend Retention Filters which control indexing.

In Datadog APM, the EXEC TIME (Execution Time) column specifically shows the duration of each span's operation. For database queries, this represents the actual time taken to execute the query. This metric directly answers the question of which database query is taking the longest time. By sorting or examining the EXEC TIME column values, you can identify the database span with the highest execution time, which is the longest-running query.

A. Check the SPANS column: Shows span names/types but not their duration or performance metrics.

C. Check the ERRORS column: Indicates error occurrences, not execution time or duration.

D. Check the LATENCY column: Represents end-to-end trace latency, not individual span execution time.

1. Datadog Documentation - "Trace View" - APM > Trace Search & Analytics section, specifically the "Span List" subsection describing column definitions

2. Datadog Documentation - "Key APM Terminology" - APM > Getting Started, defining execution time vs latency

3. Datadog Learning Center - "APM Fundamentals" course module on span analysis and performance metrics