Yandex Cloud
Search
Contact UsGet started
  • Blog
  • Pricing
  • Documentation
  • All Services
  • System Status
    • Featured
    • Infrastructure & Network
    • Data Platform
    • Containers
    • Developer tools
    • Serverless
    • Security
    • Monitoring & Resources
    • ML & AI
    • Business tools
  • All Solutions
    • By industry
    • By use case
    • Economics and Pricing
    • Security
    • Technical Support
    • Customer Stories
    • Gateway to Russia
    • Cloud for Startups
    • Education and Science
  • Blog
  • Pricing
  • Documentation
Yandex project
© 2025 Yandex.Cloud LLC
Yandex Load Testing
  • Getting started
    • Service overview
    • Agent
    • Agent selection
    • Load generator
    • Load profile
    • Testing threads
    • Autostop
    • Resource Check
    • Test agent monitoring
    • Reports
    • Load test regressions
    • Load test results
    • Comparing load test results
    • Test artifacts
    • Quotas and limits
  • Access management
  • Pricing policy
  • Terraform reference
  • Audit Trails events
  • Release notes

In this article:

  • Agent performance
  • Use cases
  1. Concepts
  2. Agent

Agent

Written by
Yandex Cloud
Updated at March 7, 2025
  • Agent performance
  • Use cases

An agent is a VM that provides computing resources to generate load on the target application. This load comes as preset test data, such as requests, parameters, or scripts.

Note that, during load generation, you need to monitor the agent system resources, such as vCPUs, memory, disks, and network. If you run out of any resource, you will not be able to generate load and measure responses in an appropriate way.

Note

Agents automatically clean up disk space. By default, an agent cleans up 2 GB of space. You can use the Resource Check plugin's disk_limit parameter to manage the amount of cleaned up space.

All agents have a load testing tool on board; they also have a connector that enables this tool to work with Load Testing.

Agent performanceAgent performance

Load Testing supports three types of agents:

  • Small: 2 vCPUs, 2GB RAM, 15GB HDD: In simple use cases, such an agent allows you to generate a load of more than 10,000 requests per second with both Phantom and Pandora. This type of agent is suitable for most load tests. If you do not need to emulate a very large request flow or special conditions (such as creating a large number of concurrent connections), you are unlikely to look any further.
  • Medium: 4 vCPUs, 4GB RAM, 30GB HDD: Allows you to establish a larger number of concurrent connections and generate a load of more than 20,000 requests per second. This type is suitable for more complex tests. You may want to choose it if the small type is not enough. With a larger storage capacity, a medium agent allows you to establish more connections or run longer tests.
  • Large: 8 vCPUs, 8GB RAM, 60GB HDD: Generates up to 40,000 requests per second. Use large agents for long tests with a large number of concurrent connections.

Specific load metrics highly depend on the test configuration. For example, you can generate more requests with the Connection: Keep-Alive header than with Connection: Close, long responses reduce the maximum possible number of requests per second, etc.

If the large agent type is not powerful enough to meet your needs, contact support.

Use casesUse cases

  • Load testing a gRPC service
  • Fixed-load HTTPS testing with Phantom
  • Step-load HTTPS testing with Pandora
  • Scripted HTTP load testing with Pandora
  • JMeter load testing
  • Load testing using multiple agents
  • Running external agents for load testing
  • Comparing load test results

Was the article helpful?

Previous
Service overview
Next
Agent selection
Yandex project
© 2025 Yandex.Cloud LLC