I just ran this script on our AWS account. This is the result I got:

│ Total Green Cost             │ Total Grey Cost              │
├──────────────────────────────┼──────────────────────────────┤
│ 100.0% ($0.08)               │ 0.0% ($0.00)                 │

Now, that's good, because we run everything in Ireland anyway, but it's bad because we definitely spent more than 8 cents last month.

Our AWS bill is fully covered by credits we got from them. It seems that the script is excluding these costs in the calculation.

I believe that a lot of companies, especially startups, have vast amounts of credits from AWS via their various accelerator programs. I think that these "costs" should be included in the calculation.

After some googling, it seems that CostExplorer.getCostAndUsage supports this via something like:

    Dimensions: {
      Key: 'RECORD_TYPE',
      Values: [
        'Credit',
      ]
    },

I'll try to get a PR up for this.

Right now, we just dump a bunch of javascript objects to the command line.

This isn't very helpful.

What would be a nice start, would be to get a lighthouse ci style breakdown like so:

What would be really nice would be to have a flag to output json in a format that works nicely with vega-lite.

This would make it straightforward to talk about analysis using observable, or if you understandably aren't crazy about sharing AWS API keys onto thirty platforms, run the checks locally.

https://beta.observablehq.com/@mbostock/exploring-data-with-vega-lite

We'd like a nice way to get releases onto npm easily.

This would include

• setting up with travis
• setting up something like np gem
• something like greenkeeper or similar to keep deps safe and er... dependable

Now we have this running, and we have an idea of what we want to support with this, we spoke about refactoring this, to make it easier to change, as the single file is doing a bunch of work:

A proposed structure

We discussed breaking this into three smaller components, like this picture here:

• CLI parser - this takes care of setting attributes based on the flags we pass in on the command line
• filter/grouper - this takes the data we care get back from cost explorer, and reshapes it so it's easy to represent in json, CLI output, on a page, and so on
• formatter - this takes the datastructure we generate with the component above, and takes care of formatting it in the chosen output format (i.e. as a json file, as html, or a terminal table)

npx greencost does not work...
npm ERR! 404 'greencost@latest' is not in the npm registry.

I just want to know if this project is still maintained. Dont want to come off as rude, but it doesnt look like that atm (no commits since 2020).

When running I'll get greeted by:

NOTE: We are formalizing our plans to enter AWS SDK for JavaScript (v2) into maintenance mode in 2023.

If we make this output in the format vegalite expects, we can then rely on VegaLite to give us all kind of nice viz, aggregations and rollups, which then makes it easy to embed in dashboards or anywhere you might use vega embed:

https://github.com/vega/vega-embed

This gives us nice PNG and PDF outputs to make it easy for people to drop into reports, emails, slack channels, etc and make the case for getting off fossil fuels.

For this, I think the key thing we need is some notion of a timestamp in each json object for the region.

Because you can npm install js in loads of places, and because you could plausibly put this on glitch or other such services, it makes sense to have some safeguard to either warn you, or not run if a set of credentials with more privileges than are strictly needed have been detected.

I don't know how to do this yet, but I'm hoping some kind soul might.

C

If you have a target of getting off fossil fuels, it helps to know which services you are spending on, as some things are easier to move across regions than others.

Also, if we had a breakdown by service, then it would be possible to show charts like this, but grouped by if they can in green vs grey regions as a way to track progress over time

This is from the post below. Worth a read.

https://blog.mapbox.com/tracking-aws-infrastructure-cost-on-a-team-level-5b84c3fc8467?gi=2e6b996a0823

When we were pairing we came across a few things that weren't obvious to debug.

Depending on the kinds of errors, we should probably:

• Make list of the common errors we've seen so far
• Make list of the common errors you see elsewhere
• Link to the most relevant pages on the AWS SDK help docs when we know them

This project needs a bunch of stuff like:

• CoC
• proper coding and contribution guidelines
• setting up with CI
• a roadmap

And more. Leaving this ticket here as a reminder (and invitation to any product people…)

Some charges have no region info attached to them or are global.

How should these be handled? Grey by default, a third "neutral" category, or disregarded entirely?

Until cloud providers provide more transparency on how much power, and therefore carbon any of the services emit, we can get some ballpark figures by either:

1. using Amazon's own emissions intensity figures from their own sustainability reporting. This is for all of Amazon, not AWS. So we'll have innaccuracies as running exclusively datacentres have a different energy use profile to all of e-commerce.

or

2. adapt this spreadsheet, and in particular the Platform part, I built when working with another agency to get some figures. For this we used the UK figures for emissions intensity of IT from the ONS, adjusted for what we knew about likely energy use, and straight up cost of energy.

As outlined in the readme, show cost data split over time using the API's given time series chunks:

|          | Grey   | Green  |
-----------|--------|--------|
|Aug  2018 |  40%   |  60%   |
|Sept 2018 |  40%   |  60%   |
|Oct  2018 |  40%   |  60%   |
|Nov  2018 |  40%   |  60%   |
|Dec  2018 |  40%   |  60%   |

rather than only

|          | Grey   | Green  |
-----------|--------|--------|
|Total     |  40%   |  60%   |

Changes are needed in two places to do this:

1. note timestamp when aggregating data by region here
2. respect timestamp when summing up total here

cc @mrchrisadams

Right now the we have hard coded dates for running checks.

There's a normal 1 year rolling window I think for cost explorer, but it would be nice to set the time bounds, like you can with the python cli.

For example, when you have this:

aws ce get-cost-and-usage --time-period Start=2018-08-01,End=2019-05-01 --granularity MONTHLY --metrics "BlendedCost" "UsageQuantity" --group-by Type=DIMENSION,Key=AZ

I would be good to do something to either use the same kinds of flags:

greencost --time-period Start=2018-08-01,End=2019-05-01

Or ideally use flags like so, which

I'm not sure what the convention is for node js, but if we know we only ever have maximum of today plus one year back, we might be able to have some sensible defaults to show the last year of spend analysis, or be able to set arbitrary figures like so:

greencost --time-period-start 2018-08-01 --time-period-end 2019-05-01

I'm not sure what the most command CLI tool for node is, but commander looks well documented, and is used by the Vue CLI, so there's code examples we can look to for seeing how different things are implemented

https://www.npmjs.com/package/commander

### Bonus bit

If filtering by tags is simple, then it gives us a nice way to restrict a query to a single project team.

https://docs.aws.amazon.com/aws-cost-management/latest/APIReference/API_GetCostAndUsage.html#API_GetCostAndUsage_RequestSyntax

TODO: fill this out after the chat