The system is composed by a manager node and a number of worker nodes.

The function to be evaluated will be defined in the worker nodes, using the selected programming language, following the following interface:

function( args: number[] ): number

The function to be evaluated will be the Grienwank function in 3 dimensions.

The data will be generated by the manager node, and will be sent to the worker nodes in batches.

The parameters will be received from a json file with the following structure:

{
  "data": [
    [start, end, step],
    ...
  ],
  "agg": "MIN" | "MAX" | "AVG",
  "maxItemsPerBatch": num
}

• data: a list of intervals where the parameters will be generated to evaluate the function.
• agg: the aggregation function to be used to aggregate the results of the function evaluation.
• maxItemsPerBatch: the maximum number of items to be sent to the worker nodes in each batch.

The intervals to be evaluated, for each of the 3 dimensions, is [-600, 600, 0.2], the aggregation function is "MIN" and the maximum number of items per batch is 10800000.

Benchmarks will be run using 4, 8 and 16 worker nodes.

• Measurements
  • Completion Time
  • CPU usage
  • Node throughput and Combined throughput
  • Memory usage
  • Network usage
  • Coefficient of variation of worktime and results produced

The metrics dashboard can be found here