Duplicate configuration.gpu.memory in aws.csv

JacobValdemar commented 9 months ago

Bug description

In aws.csv the data contains two columns with the name configuration.gpu.memory. Looking at the data, maybe the first instance should have been named to configuration.gpu.name?

https://github.com/Boavizta/boaviztapi/blob/5b6ef428cbf11e297c5c11554438d5d1b8667b19/boaviztapi/data/devices/cloud/aws.csv?plain=1#L1

To Reproduce

Expected behavior

JSON OUTPUT

Additional context

da-ekchajzer commented 9 months ago

Thanks for this bug report. We fixed it in the dev version: https://github.com/Boavizta/boaviztapi/blob/dev/boaviztapi/data/archetypes/cloud/aws.csv

The dev version should be released soon. I recommend you to use it as of today.

GPU data is not used because we don't yet have a method to take their impacts into account. Dedicated issue : https://github.com/Boavizta/boaviztapi/issues/65

JacobValdemar commented 9 months ago

@da-ekchajzer OMG I have been using main for so long and wondered why so much seemed outdated. I am happy that you shared with me that you have a dev branch. Will use that going forward 🚀

da-ekchajzer commented 9 months ago

Dev doc : dev.doc.api.boavizta.org/

Since we made some breaking changes, here is a draft of the pre-release :

v1.0.0

New features

Add new end-user devices from Base Empreinte
Add PEF impacts criteria
Add IoT device impacts
Add min/max values depending on user input completeness
Add warnings
Improvement of the completion process from CPU name
Adding around 2000 CPUs for completion
Adding utils routers (list available data for string fields, archetypes routers, etc.)
Users can now choose the impact factors to compute
Users can now add a special message to the home page of the API
Refactor and normalize routes names
CPU die is now express in mm2 instead of cm2
Refactor the allocation process based on duration

Internal changes

Facilitating the completion process
Facilitating the process of adding new devices
Facilitating the archetype process
Externalize the impact factors in a separate file
Create a config file
CPU die completion now use cpu spec file
CPU uses die_size instead of die_size_per_core

Breaking changes

Cloud routers

Before

We add one rout per cloud provider (e.g. /v1/cloud/aws)

Now

We have only one route for all cloud providers (e.g. /v1/cloud/instance)
Each route has a parameter called provider (in the url for GET requests, in the body for POST requests).

Duration & allocation

Before

Duration was a field in the usage object called hours_use_time.

{
  "usage": {
    "hours_use_time": 2,
  }
}

Allocation was a route parameter.

Now

Duration is now a route parameter. Allocation is no longer used
If not provided, we use the lifetime of the device as duration.
We compute usage impacts hover the duration and allocate embedded impacts on the duration hover the lifetime of the device.
We introduce the notion of use_time_ratio which is the proportion of time the device is used during the given duration. When a device is always used, the usage ratio is 1. When a device is never used, the usage ratio is 0.

{
  "usage": {
    "use_time_ratio": 0.5,
  }
}

Impacts format

Before

"gwp": {
      "manufacture": 1900,
      "use": 260,
      "unit": "kgCO2eq"
    },

or

"manufacture_impacts": {
        "gwp": {
          "value": 23.8,
          "unit": "kgCO2eq"
        },
        "pe": {
          "value": 353,
          "unit": "MJ"
        },
        "adp": {
          "value": 0.02,
          "unit": "kgSbeq"
        }
      }
}

Now

Impact formats are now unified.
"manufacture" is now called "embedded"

"impacts": {
   "gwp": {
     "embedded": {
       "value": 6.68,
       "min": 6.68,
       "max": 6.68,
       "warnings": [
         "End of life is not included in the calculation"
       ]
     },
     "use": "not implemented",
     "unit": "kgCO2eq",
     "description": "Total climate change"
   },

Verbose format

Before

"USAGE": {
  ...
  "usage_impacts": {
       ...
  }
  "avg_power": {
    "value": 94.62364134445255,
    "unit": "W",
    "status": "COMPLETED",
    "source": null
  },
...
}

Now

For each component, all attributes are now at the same level in the dictionary.
Usage impacts are now in the "impacts" dictionary
Usage attributes are now at the same level as the other attributes
Attributes may have a "min", a "max" and a "warnings" field.

  "CPU-1": {
      "impacts": {
        "gwp": {
          "embedded": {
            "value": 64.7,
            "significant_figures": 3,
            "min": 24.6,
            "max": 149,
            "warnings": [
              "End of life is not included in the calculation"
            ]
          },
          "use": {
            "value": 160,
            "significant_figures": 2,
            "min": 160,
            "max": 160
          },
          "unit": "kgCO2eq",
          "description": "Total climate change"
        }
      },
      "die_size_per_core": {
        "value": 0.47078947368421054,
        "status": "COMPLETED",
        "unit": "mm2",
        "source": "Average for Skylake",
        "min": 0.07,
        "max": 1.02
      }
      ... 
}

Contributors

ToDo

Known future requirements

Mobile and fix network impacts
Generalize the AWS process to other cloud providers
GPU impacts
Add multiple impact factors for depending on the engraving process size
Screen impacts from characteristics
Take into account the uncertainty of the impact factors
Adding a system layer

Boavizta / boaviztapi