Traffic Generation Mode

Overload supports two ways to specify how requests should be generated at any specific moment("second"). Can be set using generationMode field of the test request.

In conjunction with these options along with QPS & concurrent connection configuration, the generator can be configured either to open or closed system

Batch

Sends the requests in batches at certain intervals. The generator keeps track of available time to send all the required batches. If it fails to send a batch or any request in the batch, intervals might be reduced to accommodate remaining requests in the available time.

{
  "generationMode": {
    "batch": {
      "batchSize": 10
    }
  }
}

Example

Consider the generator needs to produce 100 requests in a certain second, and the batch size is 10.

In this case, the second will be sliced into ten 100ms periods. At the beginning of each slice, 10 requests will be sent to the target. If there is enough connection available in the pool, no new connection will be created, otherwise new connections will be attempted. Note that the connection creation is controlled by concurrent connection configuration.

If generator failed to send the second batch due to connection availability, slices will be recalculated. Now 9 batches needs to be sent in ~800ms. the slice duration will be updated to (100/9)=~88ms.

Time mentioned here is simplified and actual slice duration can be shorter. The generator reserves 10%(~100ms) for itself and may consume more than that.

Immediate

The default generation mode.

Tries to send all the requests at the beginning of the second. Crates as much as connection as required as long as it's allowed by concurrent connection configuration. If enough connections are not available, generator waits for next a(or more) connection to be available, and sends the request.

{
  "generationMode": "immediate"
}