System requirements

  • Reference
  • 2-min read
  • Published Nov 19, 2024

This page describes system requirements for the Dynatrace OpenTelemetry Collector Distribution for different use cases.

Hardware used

The numbers below were gathered using an Azure virtual machine of type Dadsv5-series (AMD EPYC 7763v Genoa CPU) with 4 vCPUs and 16GB RAM.

CPU and memory requirements for traces, metrics and logs (combined)

The requirements for the Dynatrace Collector are based on a load scenario with the following numbers per second:

  • 1000 traces (~1.2 KB each)
  • 1000 metrics (~1.2 KB each)
  • 1 MB logs

The recommended resources for this combined scenario are:

  • 500 MiB RAM
  • 500 mCPU

If you need additional data processing (for example, filter or transform processors), system requirements will increase.

Check out specific scaling and infrastructure architecture considerations on our separate page.

Specific load scenarios

The Dynatrace Collector was also tested under heavier scenarios than above but only with single signal types. The following table shows performance data based on the following base data sizes:

  • Traces: ~1 KiB per trace
  • Metrics: ~3 KiB per metric
  • Logs: ~3 KiB per line

This table shows throughput scenarios based on the above data sizes with different amounts of load and different used protocols.

Scenario (traces, metrics, logs per second)CPU coresRAM (MiB)
OTLP-HTTP 10k traces0.25100
OTLP-HTTP 100k traces1.5120
OTLP-HTTP 10k metrics0.25110
OTLP-HTTP 100k metrics1100
Syslog 10k logs 1 per batch0.2100
Syslog 10k logs 100 per batch0.2100
Syslog 70k logs 1 per batch1100
Syslog 70k logs 100 per batch0.5110

Prometheus scrape performance

Additional metrics based load scenarios were done based on Prometheus scraping with the following base settings:

  • 1s scrape interval
  • Metrics data size: ~3 KiB per metric

This table shows the load results with different scenarios.

ScenarioCPU coresRAM (MiB)
1 endpoint (10k data points each)0.5300
1 endpoint (1k data points each)0.1140
5 endpoints (1k data points each)1.5250
10 endpoints (1k data points each)2500
Related tags
Application Observability