Skip to main content

Weights

In ARGO you may combine top level A/R results (such as results referring to you data center, institution as a whole) into a higher level of abstraction such as organisations, projects, federation. In this level each member contributes to the final A/R score based on its resource capacity. This contribution is expressed as a weighted average, which is obtained by various ways such us by running benchmark on the infrastructure (ex. hep spec 06)

How weights are used in the computations​

ARGO Monitoring Service takes into consideration the weights to combine top level A/R results.

How can I define the weights​

In ARGO we actually define sets of weights applied to top level items (as we've mentioned, those are groups that represent data centers, institutions etc..). The definition includes:

  • a name
  • the type of top level group items that the weights are applied to
  • the type of weights (related to computational cores, benchmark scores, TFLOPS, storage capacity, etc...)
  • And the list of weighted items (tuples of item names along with the weight assigned to them which is usually a float number)

The Json definition of a weight set is described in the example below:

{
    "name": "Example_Weightset",
    "weight_type": "HEPSPEC06_score",
    "group_type": "DATACENTERS",
    "groups": [
        {
            "name": "DC_01_AMSTERDAM",
            "value": 12356.4
        },
        {
            "name": "DC_02_LONDON",
            "value": 16791.5
        },
        {
            "name": "DC_03_FRANKFURT",
            "value": 18791.5
        }
        ]
}

Monitoring Service has a number of connectors for setting the weights:

  • VAPOR
  • CSV files (predefined format)
  • Json files (predefined format)
  • xml files (predefined format)