DBM Agent Integration Overhead

이 페이지는 아직 한국어로 제공되지 않으며 번역 작업 중입니다. 번역에 관한 질문이나 의견이 있으시면 언제든지 저희에게 연락해 주십시오.

Overview

Database Monitoring runs on top of the base Datadog Agent. By default, it’s configured with optimal performance settings to minimize impact on your system. However, you have the flexibility to adjust parameters like data collection frequency and query sampling to better suit your workloads.

This page contains the results of integration overhead tests conducted against databases with Datadog Database Monitoring enabled.

Overhead testing results

    Postgres integration overhead tests were run on an Amazon EC2 machine c5.xlarge instance (4 vCPUs, 8 GB RAM). The database used for the tests was a PostgreSQL 14.10 instance running on an Amazon RDS db.m5.large instance (2 vCPUs, 8 GB RAM). The database was running a TPC-C workload with 20 warehouses.

    SettingCollection Interval
    Check Min Collection Interval15s
    Query Metrics Collection Interval10s
    Query Samples Collection Interval10s
    Settings Collection Interval600s
    Schema Collection Interval600s
    • Agent Test version: 7.50.2
    • CPU: ~1% of the CPU used on average
    • Memory: ~300 MiB of RAM used (RSS memory)
    • Network bandwidth: ~30 KB/s ▼ | 30 KB/s ▲
    • Agent query overhead on database: ~1% CPU Time

    Note: The network bandwidth is the sum of the incoming and outgoing traffic from the Agent to the monitored database and the Datadog backend.

    MySQL integration overhead tests were run on an Amazon EC2 machine c5.xlarge instance (4 vCPUs, 8 GB RAM). The database used for the tests was a MySQL 8.0 instance running on an Amazon RDS db.m5.large instance (2 vCPUs, 8 GB RAM). The database was running a TPC-C workload with 20 warehouses.

    SettingCollection Interval
    Check Min Collection Interval15s
    Query Metrics Collection Interval10s
    Query Activities Collection Interval10s
    Query Samples Collection Interval1s
    Settings Collection Interval600s
    • Agent Test version: 7.50.2
    • CPU: ~2% of the CPU used on average
    • Memory: ~300 MiB of RAM used (RSS memory)
    • Network bandwidth: ~40 KB/s ▼ | 30 KB/s ▲
    • Agent query overhead on database: ~1% CPU Time

    Note: The network bandwidth is the sum of the incoming and outgoing traffic from the Agent to the monitored database and the Datadog backend.

    SQL Server integration overhead tests were run on an Amazon EC2 machine c5.xlarge instance (4 vCPUs, 8 GB RAM). The database used for the tests was a SQL Server 2019 Standard Edition instance running on an Amazon RDS db.m5.large instance (2 vCPUs, 8 GB RAM). The database was running a TPC-C workload with 20 warehouses.

    SettingCollection Interval
    Check Min Collection Interval15s
    Query Metrics Collection Interval60s
    Query Activities Collection Interval10s
    Settings Collection Interval600s
    • Agent Test version: 7.50.2
    • CPU: ~1% of the CPU used on average
    • Memory: ~300 MiB of RAM used (RSS memory)
    • Network bandwidth: ~40 KB/s ▼ | 30 KB/s ▲
    • Agent query overhead on database: ~1% CPU Time

    Note: The network bandwidth is the sum of the incoming and outgoing traffic from the Agent to the monitored database and the Datadog backend.

    Oracle integration overhead tests were run on an Amazon EC2 machine c5.xlarge instance (4 vCPUs, 8 GB RAM). The database used for the tests was a Oracle 19c instance running on an Amazon RDS db.m5.large instance (2 vCPUs, 8 GB RAM). The database was running a TPC-C workload with 20 warehouses.

    SettingCollection Interval
    Check Min Collection Interval10s
    Query Metrics Collection Interval60s
    Query Activities Collection Interval10s
    • Agent Test version: 7.53.0
    • CPU: ~0.2% of the CPU used on average
    • Memory: ~270 MiB of RAM used (RSS memory)
    • Network bandwidth: ~6 KB/s ▼ | 4 KB/s ▲
    • Agent query overhead on database: ~0.2% CPU Time

    Note: The network bandwidth is the sum of the incoming and outgoing traffic from the Agent to the monitored database and the Datadog backend.

    PREVIEWING: may/unit-testing