Micromesh

This project is for a simple web micro service proxy that will log and route web service requests. The project is based on .NET Core.

Quick usage

The service comes pre-configured with DuckDuck and Postman services, so the easiest ways to test it is to run the mesh via one of the run scripts ./scripts/run.ps1 or ./scripts/run.sh and test it by sending a CURL request to post man.

Windows:

curl -uri http://127.0.0.1:54492/get?foo=bar -headers @{forward-to="postman"}

Mac:

curl -url http://127.0.0.1:54492/get?foo=bar -header "forward-to: postman"

Configuration

The project is built with a goal of being as lightweight as possible both on resources, maintenance and cognitive load.

For logging you can use the built in logging of .NET Core - if you know how to configure that log, you are good.

For service configuration all you need to do is add sections to appsettings.json. Each service has a pool of one or more pre-configured hosts to be selected at random when proxying requests.

  "Services": {
    "duckduck": [ 
      "http://api.duckduckgo.com:80",
      "https://api.duckduckgo.com:443",
      "api.duckduckgo.com"
    ],
    "postman": [ 
      "postman-echo.com"
    ]
  }
  "ConversationLogging": {
    "LoggingEnabled" : "true",
    "IncludedServices": [
      "foobar",
      "foobar-document"
    ]
  },
  "RetryPolicy": {
    "ExceptionRetryCount": "3",
    "StatusCodeRetryCounts": {
      "404": "1",
      "408": "1",
      "500": "1",
      "503": "3"
    }
  }

Routing traffic through the mesh.

You can run the mesh server on any port, and the clients of the mesh should direct the traffic to that IP/URL/Port with a host header that points to one of the configuration items.

Performance Tests

We tested Micromesh using two scenarios as follows:
Scenario 1: Sending 10,000 requests with a small JSON payload.
Scenario 2: Sending 1,000 requests with a payload containing an 827 Kb PDF file.
Below are the results for average request time.