COOL — Cooling Operations Optimization Layer

The Cooling Operations Optimization Layer (COOL) answers one question: given a facility’s cooling technology, age, and climate, how does its Power Usage Effectiveness compare with the documented benchmark? A PUE of 1.45 can be reasonable for an older air-cooled building in a hot climate and weak for a newer liquid-cooled building in a cooler climate. COOL provides that context through a benchmark matrix derived from public survey data, adjusted for ASHRAE climate zone, and produces a letter grade (A-F) plus a continuous efficiency score consumed by GRADE for assessment scoring.

Primary metric

cool_efficiency_score — a continuous 0.0–1.0 score derived from the gap between your reported PUE and the climate-adjusted efficient benchmark for your facility type.

delta = annual_pue - climate_adjusted_benchmark_pue

cool_efficiency_score:
  delta ≤ -0.05  → 1.0   (outperforming benchmark by 5%+)
  delta =  0.00  → 0.857 (exactly at benchmark)
  delta ≥  0.30  → 0.0   (failing threshold)

The score is a linear interpolation of delta across the interval [-0.05, +0.30]. Values outside that interval are clamped. Every hundredth of a PUE point matters.

Letter grade

COOL assigns a letter grade based on the same delta:

Grade	Condition	Label
A	delta ≤ -0.05	Excellent
B	-0.05 < delta ≤ 0.05	Good
C	0.05 < delta ≤ 0.15	Adequate
D	0.15 < delta ≤ 0.30	Poor
F	delta > 0.30	Failing

Benchmark methodology

The benchmark varies by three dimensions: cooling type, facility age band, and ASHRAE climate zone.

Cooling types: air, water, liquid_to_chip, hybrid

Age bands:

Band	Years
`pre_2010`	Built or last substantially renovated before 2010
`2010_2018`	2010–2018
`post_2018`	After 2018

Each combination has a typical benchmark (published survey median) and an efficient benchmark (top quartile). COOL grades against the efficient benchmark — the bar is set by what the best-performing facilities of the same type report, not the average.

Example benchmarks (efficient, pre-climate-adjustment):

Cooling type	pre_2010	2010_2018	post_2018
air	1.55	1.40	1.30
water	1.45	1.30	1.20
liquid_to_chip	1.35	1.20	1.10
hybrid	1.45	1.28	1.15

Sources: Uptime Institute Global Data Center Survey 2023, Green Grid Data Center Efficiency Metrics 2022, DOE Data Center Optimization Initiative 2022.

Climate adjustment

Facilities in hotter climates face a larger cooling burden. COOL adjusts the efficient benchmark by an additive offset derived from the facility’s ASHRAE climate zone before comparing against reported PUE.

climate_adjusted_benchmark = efficient_benchmark + climate_adjustment

Positive adjustment = benchmark relaxed (hotter climate, more cooling load). Negative adjustment = benchmark tightened (colder climate, free cooling available).

ASHRAE Zone	Name	Adjustment	Examples
1A	Very Hot Humid	+0.10	Miami FL, Houston TX
2A	Hot Humid	+0.07	Atlanta GA, Dallas TX
2B	Hot Dry	+0.05	Phoenix AZ
3A	Warm Humid	+0.04	Charlotte NC
3B	Warm Dry	+0.02	Los Angeles CA
3C/4A	Warm Marine / Mixed Humid	0.00	San Francisco CA, Baltimore MD
4C	Mixed Marine	-0.03	Seattle WA, Portland OR
5A	Cool Humid	-0.04	Chicago IL, Detroit MI
5B	Cool Dry	-0.05	Denver CO, Salt Lake City UT
6A	Cold Humid	-0.07	Burlington VT, Duluth MN
6B	Cold Dry	-0.08	Helena MT
7	Very Cold	-0.10	International Falls MN
8	Subarctic/Arctic	-0.12	Anchorage AK

Source: ASHRAE Thermal Guidelines for Data Processing Environments, 5th Edition.

For US facilities, COOL maps the state abbreviation to the predominant ASHRAE zone for that state. For international facilities, supply the ASHRAE zone code directly — ASHRAE zones are globally defined.

Worked example

Facility: Air-cooled, built 2008, Minneapolis MN, annual PUE 1.72

cooling_type: air
age_band:     pre_2010  (built 2008)
location:     MN → ASHRAE zone 6A (Cold Humid) → adjustment -0.07

efficient_benchmark:          1.55  (air, pre_2010)
climate_adjusted_benchmark:   1.55 + (-0.07) = 1.48

delta = 1.72 - 1.48 = +0.24

Grade: D (Poor)  — 0.15 < delta ≤ 0.30
cool_efficiency_score: (0.30 - 0.24) / 0.35 = 0.171

Despite operating in a cold climate (which tightens the benchmark), the facility’s old air cooling is pulling it down to D.

Contrast — Liquid-cooled, built 2021, Houston TX, annual PUE 1.08

cooling_type: liquid_to_chip
age_band:     post_2018  (built 2021)
location:     TX → zone 2A (Hot Humid) → adjustment +0.07

efficient_benchmark:          1.10  (liquid_to_chip, post_2018)
climate_adjusted_benchmark:   1.10 + 0.07 = 1.17

delta = 1.08 - 1.17 = -0.09

Grade: A (Excellent)  — delta ≤ -0.05
cool_efficiency_score: 1.0  (clamped)

Modern liquid cooling in a hot climate, outperforming an already-relaxed benchmark.

International support

For facilities outside the US, provide the ASHRAE zone code directly in the input. ASHRAE climate zones are globally defined:

Location	ASHRAE Zone
Netherlands	5A
Singapore	1A
Norway	7
China (Beijing)	5A
China (Shenzhen)	2A

Findings exported to GRADE

Metric	Description
`cool_efficiency_score`	Continuous 0.0–1.0 score — GRADE input
`pue_reported`	Organization self-reported annual average PUE
`pue_benchmark`	Climate-adjusted efficient benchmark
`pue_gap`	`pue_reported` minus `pue_benchmark` (the delta)
`cooling_grade`	Letter grade as 0–4 integer (A=4, B=3, C=2, D=1, F=0)
`climate_adjustment_applied`	Additive ASHRAE zone offset applied to benchmark

All findings carry confidence: medium for self-reported Level 1 input. Computed values (benchmark, climate adjustment) carry confidence: high.

Confidence and self-reporting

PUE at Level 1 is self-reported. PTL does not independently verify the value. The grade reflects what the organization reports, so the assessment is only as reliable as the input.

Level 2 (BMS timeseries from a Building Management System) is specified but not yet implemented. When available, it will enable seasonal variance analysis and higher-confidence findings.

Known limits

Climate zone lookup is state-level for US facilities. Organizations in large, climatically varied states (California, Texas) may be assigned a zone that does not reflect their specific location. International facilities should supply the zone code directly.
Benchmarks reflect public survey data through 2023 and will be reviewed annually.
COOL v0.1.0 does not adjust for compute density. A high-density GPU cluster has higher heat rejection requirements than a traditional HPC cluster at the same PUE — this is not yet captured.

Input schema

Field	Type	Required	Description
`organization`	string	yes	Organization name
`period`	string	yes	Reporting year (YYYY)
`cooling_type`	string	yes	`air`, `water`, `liquid_to_chip`, or `hybrid`
`facility_age`	int	yes	Year built or last substantially renovated
`location_city`	string	yes	City name (for display)
`location_state`	string	yes	2-letter US state abbreviation, or `""` for international
`annual_pue`	float	yes	Annual average Power Usage Effectiveness
`ashrae_zone`	string	no	Direct ASHRAE zone override — required for international facilities

CLI usage

# Score a single facility from a cool_input_v1.json file
cool analyze \
  --input cool_input.json \
  --output cool_output.json

# Run all five synthetic organizations and print summary table
cool demo

# Validate a ptl_output_v1.json file against the PTL schema
cool validate --input cool_output.json

Example cool_input.json:

{
  "organization": "Midwest University HPC Center",
  "period": "2026",
  "cooling_type": "air",
  "facility_age": 2008,
  "location_city": "Minneapolis",
  "location_state": "MN",
  "annual_pue": 1.72
}

International facility:

{
  "organization": "Netherlands AI Institute",
  "period": "2026",
  "cooling_type": "water",
  "facility_age": 2019,
  "location_city": "Amsterdam",
  "location_state": "",
  "annual_pue": 1.25,
  "ashrae_zone": "5A"
}