The function (sequence) is to find the n-th positive number.

It is implemented in 2 ways:

• Nested Sequence

  The action invokes another action, and returns until the invoked action returns.

  n can be assigned to any positive integer theoretically.

  However, if n is too large, it's likely to get a timeout as the first action needs to receive the result in 60s. It usually depends on the cold boot time of a runtime container

  Besides, n should not be greater than the maximum allowed number of the concurrent runtime containers, which is limited by openwhisk.

• Chained Sequence

  The action is implemented by the "action sequence" provided by Openwhisk. Each action delivers its result to another action and finish itself. The number of action is statically assigned to 100, and each action makes the result += 1.

The 2 functions are designed to compare these 2 ways to implement an action sequence.

• Nested Sequence

  • n

    Should be int and positive.

• Chained Sequence

  • n

    Should be 0 to get the final result 100.

• Nested Sequence

  • The start time of each action (timestamp; millisecond)
  • The return time of each action (timestamp; millisecond)
  • The time of each action to invoke another action, except the last one(timestamp; millisecond)
  • the result (n)

• Chained Sequence

  • The start time of each action
  • The return time of each action
  • The result (n)

Run ./action_update.sh in the ./action/

Run ./action_invoke.sh or ./action_invoke_seq.sh in the ./action/

• single-cold_warm.sh

  Usage:

  ./single-code_warm [-m <mode>] [-t <loop times>] [-r <result file>] [-w <warm up times>] [-l] 

  Invoke the action one by one, and echo the invoking timestamp and the returning timestamp.

  • mode: can be "warm" or "cold".

    The default mode is "warm"

    • warm

      The action invokes the first action to ensure that there's a paused container, and then invoke <loop times> actions.

    • cold

      Invoke <loop times> actions. Before every action is invoked, stop all the paused runtime containers of the action first.

  • loop times

    The times to invoke the action.

    If the mode is "warm", the default value is 10;

    else, the default value is 3.

  • warm up times

    If the mode is "warm", the action needs to be warmed up. So, the action will firstly be invoked <warm up times> times.

    The default value is 1.

  • result file

    With the specified file directory, the final result of the test will be appended to the end of the file.

  • -l

    Output the timestamps to the log.

  • -W

    Warm up only: The script only warms up the container and then stops.

    '-R' and '-W' can not be assigned at the same time. -W could only be used in 'warm' mode

  • -R

    Run only: The script does not warm up the container.

    This mode requires existing paused containers of this action.

    '-R' and '-W' can not be assigned at the same time.

• run.py

  Usage:

  python3 run.py <client number> <looptimes> [<warm up times>]

  The python script let <client number> clients invoke the action concurrently, and each client invokes the action <looptimes> times.

  Output

  • result.csv

    The invoking timestamp and the returning timestamp of each activation.

  • stdout

    The latency and the throughput.

  • eval-result.log

    Append the results to the log.

• eval.sh

  Usage:

  ./eval.sh

  Run several tests and output a comprehensive result.

  Result log: eval-result.log

Because of the high probability of timeout, this way to implement an arraySum function is not recommended. As a result, no general test script is provided.