Skip to content

Latest commit

 

History

History
166 lines (124 loc) · 3.99 KB

README.md

File metadata and controls

166 lines (124 loc) · 3.99 KB
Raw

echodelaysvc

The echodelaysvc is a service used for simulating a downstream service with configurable response time characteristics. The response times are calibrated to be as close as the desired characteristics as possible. This is useful in resiliency test situations where downstream services become faulty or non-responsive. A resilient service will be able to serve the same amount of work without toppling even when the downstream services get slower.

API

Controlled-Time Echo

Request: GET /echodelaysvc/echo/{echostring}

Response:

{
  "path" : "{echostring}",
  "planned-delay" : 1726,
  "real-delay" : 1717
}

Synopsis:

Echos the {echostring} path element back as part of a JSON response. In addition, the injected delay and the actual resulting delay are provided as part of the response in the field "planned-delay" and "real-delay", consequently.

Example:

GET /echodelaysvc/echo/foo
{
  "path" : "foo",
  "planned-delay" : 1726,
  "real-delay" : 1717
}

Set Time Distribution to Negative Exponential

Request: GET /echodelaysvc/delay/ne?min={mintime}&mean={meantime}&max={maxtime}&truncate={trunc}

Response:

{
  "type" : "NegativeExponential",
  "min" : "{mintime} {unit}",
  "mean" : "{meantime} {unit}",
  "max" : "{maxtime} {unit}",
  "truncate" : {true/false}
}

Synopsis:

Sets the response time characteristics to follow a negative exponential distribution with the given min, mean, and max time. The truncate parameter is a boolean, optional, and defaults to false. When truncate is set to false, the distribution is shifted to start at the given min, which should be the theoretical lower limit of the response time for the service it should emulate. The curve continues naturally up to max. All sampled response times above max are set to max. When truncate is set to true, the distribution starts at 0. Any sampled response times below min are set to min. All sampled response times above max are set to max. In this mode, a large number of the lower sampled response times may be set to min as a result of the lower truncation.

Example:

GET /echodelaysvc/delay/ne?min=50ms&mean=200ms&max=2s
{
  "type" : "NegativeExponential",
  "min" : "50 milliseconds",
  "mean" : "200 milliseconds",
  "max" : "2 seconds",
  "truncate" : false
}

Set Time Distribution to Gaussian

Request: GET echodelaysvc/delay/gaussian?min={mintime}&mean={meantime}&max={maxtime}&sigma={sigmatime}

Response:

{
  "type" : "Gaussian",
  "min" : "{mintime} {unit}",
  "mean" : "{meantime} {unit}",
  "max" : "{maxtime} {unit}",
  "sigma" : "{sigmatime} {unit}"
}

Synopsis:

Sets the response time characteristics to follow a gaussian distribution with the given min, mean, and max time, and the standard deviation sigma. Sampled response times below min are set to min. Sampled response times above max are set to max.

Example:

GET /echodelaysvc/delay/gaussian?min=50ms&mean=1s&max=2s&sigma=330ms
{
  "type" : "Gaussian",
  "min" : "50 milliseconds",
  "mean" : "1 second",
  "max" : "2 seconds",
  "sigma" : "330 milliseconds"
}

Set Time Distribution to Uniform

Request: GET /echodelaysvc/delay/uniform?min={mintime}&max={maxtime}

Response:

{
  "type" : "Uniform",
  "min" : "{mintime} {unit}",
  "max" : "{maxtime} {unit}"
}

Synopsis:

Sets the response time characteristics to follow a uniform distribution between the given min and max time. Since this is a uniform distribution, no truncation of sampled times are needed.

Example:

GET /echodelaysvc/delay/uniform?min=30ms&max=1s
{
  "type" : "Uniform",
  "min" : "30 milliseconds",
  "max" : "1 second"
}

Get the Current Delay Compensation Time

Request: GET /echodelaysvc/delay/compensate

Response:

{
  "total-compensate" : "{time} ms"
}

Synopsis:

Obtains the current delay compensation time, in milliseconds

Example:

{
  "total-compensate" : "5.678 ms"
}