이 페이지는 아직 한국어로 제공되지 않습니다. 번역 작업 중입니다. 현재 번역 프로젝트에 대한 질문이나 피드백이 있으신 경우 언제든지 연락주시기 바랍니다.
APM Recommendations help you improve your applications’ performance and reliability by surfacing optimization opportunities from your collected telemetry. These recommendations are designed to:
Identify and resolve performance bottlenecks
Improve service reliability and uptime
Improve end-user experience
Join the AI Recommendations Preview!
AI-driven recommendation types are now available, expanding the set of optimization opportunities Datadog can detect.
Certain recommendations rely on specific Datadog products. Use the Recommendation Prerequisite dropdown to filter recommendations by the Datadog products in your setup.
Recommendations are based on data collected from different parts of your stack:
Distributed traces from Application Performance Monitoring (APM)
Database telemetry from Database Monitoring (DBM)
Sessions and user journeys from Real User Monitoring (RUM)
Datadog correlates these sources to identify opportunities to improve performance, reliability, and user experience.
Datadog ranks recommendations by computing a priority score that weighs the potential impact of an issue against telemetry signals, such as relative request volume and performance trends. The most critical insights for improving service reliability and performance appear first.
Using recommendations
To review recommendations that need your attention:
Select a recommendation from the list to see a detailed description of the issue.
Review the problem, impact, and Datadog’s recommendation for resolving it.
(Optional) To use Bits Code to generate a code fix, under Next Steps, click Fix with Bits.
(Optional) To track the fix in Jira or Case Management, under Triage, click Add Jira Ticket or Add Case.
After you’ve reviewed the recommendation, you can use the FOR REVIEW dropdown to change the recommendation status to REVIEWED, IGNORED, or RESOLVED.
Supported recommendations
Recommendation Category
Recommendation Type
Scope of Recommendation
Recommendation Description
Recommendation Prerequisite
Performance
N+1 Queries on Database
Backend services
A backend application calls the same database sequentially instead of batching queries.
APM
Performance
Repeated Sequential API calls
Backend services
A backend application makes multiple calls to the same downstream API sequentially instead of executing them in parallel, unnecessarily increasing request latency and slowing overall service performance.
APM
Performance
Persistent Retries
Backend services
A backend application issues an excessive number of retry attempts when calling a downstream API, extending request duration and risking cascading failures under strain.
APM
Performance
Missing index
Databases
The query’s execution plan performs expensive sequential scans. When detected, Datadog recommends using an index to expedite the query.
APM + DBM
Performance
Unbalanced Read Load
Databases
A service is making read-only queries to a primary database instance when replicas are available. Routing these queries to replicas can reduce primary load and improve performance.
APM + DBM
Reliability
Aggressive Retries
Backend services
A backend application triggers rapid-fire retry attempts without adequate backoff, sustaining high pressure on struggling dependencies and risking prolonged outages by preventing system recovery during transient failures.
APM
Reliability
High Exception Volumes
Backend services
A backend application is throwing a high number of exceptions as control-flow, adding CPU and memory overhead.
APM + Continuous Profiler
Reliability
Dependency Timeouts
Backend services
A backend application times out while calling a downstream dependency because the dependency responds too slowly, causing request failures that impact end users and increase the risk of cascading failures upstream.
APM + RUM
Performance
Missing Cache
Backend services
A service performs expensive, repeated work on the request path that could be served from a short-lived cache, reducing tail latency and downstream load.
APM + AI Recs (Preview)
Performance
Tail Latency
Backend services
A service exhibits extreme tail latency driven by slow downstream spans on the critical path, often from unbounded dependency latency or sequential calls that could run concurrently.
APM + AI Recs (Preview)
Performance
Excessive Serialization
Backend services
A service spends a significant share of request time on CPU-bound serialization or parsing work, adding avoidable latency and CPU overhead.
APM + AI Recs (Preview)
Performance
Unbounded Payload
Backend services
A service accepts request parameters without size or range bounds, allowing oversized inputs to drive expensive downstream work, tail latency, and timeouts.
APM + AI Recs (Preview)
Performance
Resource Contention
Backend services
Request handling is serialized behind a synchronization primitive or long-running critical section, causing tail latency under concurrency.
APM + AI Recs (Preview)
Reliability
Connection Pool Exhaustion
Backend services
A service repeatedly exhausts its connection pool to a downstream dependency, queueing requests and causing latency spikes or failures under load.
APM + AI Recs (Preview)
Reliability
Error Misclassification
Backend services
A service surfaces expected outcomes as errors in APM, inflating endpoint error rates and obscuring real reliability regressions.
APM + AI Recs (Preview)
Note: If you use both APM and Database Monitoring (DBM), you may see fewer Missing Index recommendations here than on the DBM Recommendations page. APM Recommendations only surface Missing Index issues that Datadog can associate with an instrumented application service. Missing Index recommendations that cannot be linked to a specific service appear only in DBM.
