DCEP cheat sheet

Assessment scope: PUE boundary must be defined before measurement - total facility energy (utility meter) divided by IT equipment energy (PDU/rack meters), with consistent time windows for both.

Measurement-grade meters for PUE should be Class 0.5 or better (revenue-grade) on the utility side; branch-circuit or rack PDUs at 1% accuracy or better for IT loads.

Sampling interval rule: 15-minute or finer logging is required to resolve diurnal swings; monthly utility bills alone cannot support PUE reporting.

Energy end-use breakdown typically splits into IT load, cooling (chillers, CRAHs, pumps, towers), UPS losses, lighting, and other - cooling is usually the largest non-IT category.

Assessment sequence: establish boundary → inventory loads → meter/verify → compute metrics → identify savings → verify savings (M&V per IPMVP options A/B/C/D).

Baseline normalization: adjust for IT load, weather (CDD/HDD), and occupancy before comparing periods; raw PUE comparisons across seasons are invalid.

Free cooling hours, chiller plant kW/ton, and air-side economizer availability are the highest-yield assessment targets in most facilities.

Data quality tiers: measured (permanent metering) > temporary metering > calculated from nameplate > estimated - report the tier with every metric.

Power proportionality: server power is roughly P = P_idle + (P_max − P_idle) × utilization; idle power is often 30–50% of max, so low utilization wastes energy.

IT load consolidation decision rule: consolidate workloads onto the fewest servers that keep peak utilization below ~70–80% to preserve headroom and avoid queuing latency.

Nameplate ratings overstate real draw; use measured or vendor power-curve data (e.g., SPECpower or manufacturer power calculators) for planning.

Virtualization and container density reduce idle-server count; each decommissioned server removes both its IT draw and its share of cooling/UPS overhead (multiply by PUE).

Power capping (RAPL, BMC-level) trades performance for a hard power ceiling; useful for capping peak demand charges and matching rack power budgets.

CPU frequency/voltage scaling and C-states reduce power at low load; disabling power management in BIOS forfeits substantial idle savings.

Workload placement should consider both server efficiency and thermal/electrical topology - placing load on hot racks or near capacity raises cooling energy.

Track server utilization distribution (e.g., % of servers below 20% utilization); fleets with many idle servers have the largest consolidation upside.

Air-side efficiency: reduce bypass and recirculation by sealing cable openings, blanking unused U-space, and installing containment (hot-aisle or cold-aisle).

ASHRAE TC 9.9 recommended envelope: supply air 18–27 °C (64.4–80.6 °F) for Class A1/A2; widening within the allowable envelope enables economizer hours.

Chiller efficiency is expressed as kW/ton or COP; centrifugal chillers commonly 0.5–0.7 kW/ton at design, with part-load performance governed by the IPLV/NPLV curve.

Cooling tower approach temperature (leaving water temp minus wet-bulb) drives water-side economizer viability; lower approach = more free cooling hours.

Airflow rule of thumb: ~100–160 CFM per kW of IT load for typical raised-floor designs; verify with CFD or airflow measurements rather than assuming.

Return air temperature (RAT) vs supply (SAT) delta-T below design (e.g., < 8–10 °C) signals bypass or low load, not necessarily excess cooling.

Chilled water supply temperature reset upward (e.g., 7 °C to 12–15 °C) increases chiller and economizer efficiency but must respect coil and dew-point limits.

Fan power scales roughly with the cube of speed (affinity laws), so VFD-driven CRAH/air-handler fans at part load save substantially.

Electrical chain efficiency: utility transformer → switchgear → UPS → PDU → rack PDU → IT; each stage has losses that multiply, so overall distribution efficiency is the product of stage efficiencies.

UPS efficiency is load-dependent: modern double-conversion units reach 94–96% at high load but drop sharply below ~30–40% load; ECO/eco-mode bypasses conversion for ~98–99%.

Transformer losses split into no-load (core/iron) and load-dependent (copper) losses; lightly loaded transformers waste core loss continuously.

Power factor correction reduces reactive current and line losses; low PF also triggers utility penalties and reduces available capacity.

3-phase power: P(kW) = √3 × V_LL × I × PF / 1000; use line-to-line voltage with the √3 form, line-to-neutral with the 3× form.

Right-size UPS modules and use modular/scalable UPS so modules run near their efficient band; avoid N+1 modules idling at very low load.

Busway and higher distribution voltages (e.g., 415/480 V vs 208 V) reduce I²R losses for the same delivered power.

Redundancy topology (N, N+1, 2N) affects efficiency: 2N systems typically run at half load and therefore lower efficiency than N+1 at the same IT load.

PUE = Total Facility Energy ÷ IT Equipment Energy; ideal is 1.0; DCiE = 1/PUE expressed as a percentage.

pPUE (partial PUE) applies the ratio to a defined subsystem (e.g., one module or hall) - useful when full-facility metering is unavailable, but not comparable to facility PUE.

Other metrics: CUE (total CO2 ÷ IT energy), WUE (water liters ÷ IT kWh), ERE (Energy Reuse Effectiveness), and ITUE (IT Energy Efficiency).

ERE = (Total Energy − Reused Energy) ÷ IT Energy; values below 1.0 indicate heat reuse is offsetting facility load.

Benchmarking requires matched conditions: same boundary, same measurement method, and weather/load normalization; cross-facility PUE comparisons without normalization are misleading.

Reporting cadence: PUE should be reported as a time-weighted average over the period (energy ratio), never as an average of instantaneous PUE samples.

Annualized PUE should state the measurement period, meter accuracy class, and whether renewable energy certificates or offsets were applied.

Green Grid/ISO/IEC 30134 series defines the standard metric formulas; cite the standard version when reporting to avoid definition drift.

PUE = Total Facility Energy ÷ IT Equipment Energy; ideal is 1.0; DCiE = 1/PUE expressed as a percentage.

pPUE (partial PUE) applies the ratio to a defined subsystem (e.g., one module or hall) - useful when full-facility metering is unavailable, but not comparable to facility PUE.

Other metrics: CUE (total CO2 ÷ IT energy), WUE (water liters ÷ IT kWh), ERE (Energy Reuse Effectiveness), and ITUE (IT Energy Efficiency).

ERE = (Total Energy − Reused Energy) ÷ IT Energy; values below 1.0 indicate heat reuse is offsetting facility load.

Benchmarking requires matched conditions: same boundary, same measurement method, and weather/load normalization; cross-facility PUE comparisons without normalization are misleading.

Reporting cadence: PUE should be reported as a time-weighted average over the period (energy ratio), never as an average of instantaneous PUE samples.

Annualized PUE should state the measurement period, meter accuracy class, and whether renewable energy certificates or offsets were applied.

Green Grid/ISO/IEC 30134 series defines the standard metric formulas; cite the standard version when reporting to avoid definition drift.