invoker

Invoker

The invoker is a tool commonly used for calling and testing deployed functions / benchmarks. It reads the target addresses specified inside the input file (endpoints.json by default), and then writes the measured latencies to an output file (rps<RPS>_lat.csv by default, where <RPS> is the observed requests-per-sec value) for further analysis.

There are runtime arguments (e.g., RPS or requests-per-second target, experiment duration) that you can specify if necessary.

Input file

The addresses to be invoked are stored in an input json file, named endpoints.json by default. The standard structure of this file is a list of json objects which have a hostname. In the event that more than one object (i.e. more than one hostname) is included, then the provided addresses will be called by the invoker in a round-robin fashion.

An example input file can look like this:

[
  {
    "hostname": "producer.chained-functions-serving.192.168.1.240.sslip.io"
  }
]

For knative eventing, the invokation of functions with the goal of measuring their latency is more difficult since the connection does not stay open until a response arrives. For this reason we have developed TimeseriesDB. To use the invoker with TimeseriesDB the input json object will need an additional matchers field, which is a mapping of CloudEvent attribute names and values that are sought in completion events to exist. Please refer to the benchmarking methodology documentation for additional information.

An example input file for eventing can look like this:

[
  {
    "hostname": "producer.chained-functions-eventing.192.168.1.240.sslip.io",
    "eventing": true,
    "matchers": {
      "type": "greeting",
      "source": "consumer"
    }
  }
]

Invoker usage

make invoker can be used to build a binary of the invoker. Once this and the input file are ready, calling ./invoker will start the process.

Note: You need to have go and protoc installed on your machine to build the invoker. Refer to this for installation instructions or our install scripts ./utils/install_go.sh and ./utils/install_protoc.sh.

Additional inputs can be provided to set the desired duration (--time), requests per second (-rps) or port (-port) to be used. -dbg can be set for extra debug logs, and additionally tracing can be enabled for the invoker using the boolean -trace flag, in which case the -zipkin option may also need to be used in order to specify the address of the zipkin span collector.

Example usage:

$> ./invoker -port 31080 -dbg
time="2021-08-24T14:47:47.408226132Z" level=debug msg="Debug logging is enabled"
time="2021-08-24T14:47:47.408323683Z" level=info msg="Reading the endpoints from the file: endpoints.json"
time="2021-08-24T14:47:47.408517689Z" level=debug msg="Invoking: $HOSTNAME:31080"
time="2021-08-24T14:47:48.408726119Z" level=debug msg="Invoking: $HOSTNAME:31080"
time="2021-08-24T14:47:49.408669362Z" level=debug msg="Invoking: $HOSTNAME:31080"
time="2021-08-24T14:47:50.409601652Z" level=debug msg="Invoking: $HOSTNAME:31080"
time="2021-08-24T14:47:51.409282809Z" level=debug msg="Invoking: $HOSTNAME:31080"
time="2021-08-24T14:47:52.409379749Z" level=info msg="Issued / completed requests: 5, 0"
time="2021-08-24T14:47:52.409461172Z" level=info msg="Real / target RPS: 0.60 / 1"
time="2021-08-24T14:47:52.409479751Z" level=info msg="Experiment finished!"
time="2021-08-24T14:47:52.409522576Z" level=info msg="The measured latencies are saved in rps0.60_lat.csv"