kyhau / curl-performance-test-tool

A simple tool to benchmark http latency with curl, running in AWS Lambda Function and publish CloudWatch Custom Metrics and Dashboard.

curl-performance-test-tool

A simple tool to benchmark http latency with curl.

For details of the timings see Cheat Sheet on Curl Performance Metrics: how to benchmark server latency with curl.

---

The deploy-curl-performance-test-master workflow

• Deploy a AWS Lambda Function that runs curl on the given list of URLs regularly (every 15 minutes by default).
• Produce the following set of CloudWatch Custom Metrics under the a custom Namespace (e.g. CurlPerformance.
  1. DnsLookUpTimeInMs
  2. TcpConnectionTimeInMs
  3. TlsHandshakeTimeInMs
  4. RequestSentTimeInMs
  5. FirstByteTimeInMs
  6. ContentTransferTimeInMs
  7. TotalDurationInMs
  8. Failed (unit: None)
• Produce a CloudWatch Dashboard named CurlPerformance.

Specifying URLs to be tested

• The upload-input-files-to-s3 workflow uploads the URL input files in inputs/ to a S3 bucket which is read by the Lambda Function.

• URLs are specified in the corresponding inputs/<AccountId>-<Region>.json file.

• In the example below, github_com is used for the CloudWatch Metric Dimension Id in the Namespace CurlPerformance.

  {
    "github_com": {
      "Url": "https://github.com"
    }
  }

• To add URLs

  1. Add the URLs to the corresponding inputs/<AccountId>-<Region>.json.
  2. Merge the change.
  3. The workflow will be triggered automatically to upload the file to the S3 bucket and recreate the CloudWatch Dashboard.

To deploy a new stack to a new AWS account or a new Region

1. Edit .github/workflows/deploy-curl-performance-test.yaml - add VPC ID and Subnet IDs to the step with id VpcInfo.
2. Edit .github/workflows/upload-input-files-to-s3.yaml - add the aws account alias to accountenvironment.
3. Create inputs/<AccountId>-<Region>.json.
4. Merge the change.
5. Trigger the workflow deploy-curl-performance-test-master.

TODOs

1. The Lambda Function uses Python 3.7 Runtime which has curl installed in its Amazon Linux 1 environment. Some work is required to update the Runtime to Python 3.8 or 3.9 which do not have curl installed in the Amazon Linux 2 environment.