CSCI PSU Proposal for ENERGY STAR Storage Server Specification€¦ · CSCI PSU Proposal for ENERGY...

CSCI PSU Proposal for ENERGY STAR Storage Server Specification

EPA Energy Star for Storage Development Workshop

Kathleen FiehrerCSCI Technical Program Manager

Energy Storage Spec Framework Document

• Power Supply Requirements: EPA stated intentions of exploring a Net Power Loss approach for Computer Storage and Tier 2 Server power supplies. • Approach would aim to specify a maximum allowed

power loss through the power supply at actual operating conditions (e.g., Idle and full load power).

• Alternatives •Specify the minimum efficiency for a power supply under actual use conditions

•Evaluate both Multi-Output and Single-Output power supply efficiency and power factor levels.

ENERGY STAR Motivation for NPL Approach

• ENERGY STAR Goals for Storage PSU requirements • Maximize and promote operational efficiency • Encourage right sizing and address impact of redundancy • Continue to recognize PSU as possible energy bottleneck AND

ALSO opportunity for savings regardless of hardware configuration, work load, or application

• Maintain momentum toward more efficient supplies as mainstream

• Current method does not consider how a PSU is installed and operated in the field.

ENERGY STAR NPL Method • ENERGY STAR has indicated that NPL is a possible

approach for measuring system efficiency rather than current PSU efficiency and PF approach • Measures AC Powerin – DC Powerout at Idle and Max vs current

method DC Powerout / AC Powerin at designated rated loads • Address actual wattage losses at real operating conditions

instead of efficiency at arbitrary load conditions • Directly correlates to wasted wattage / real-world conditions • Eliminate current arbitrary load points of 10%, 20%, 50% and 100%

loading • Includes effects of redundancy and PSU sizing

• Industry opposes • No accepted protocol • Industry aligned behind current methods

Gold PSU Losses vs Losses 0 0.2 0.4 0.6 0.8 1

% Load

180 94%

160 92%

Max

Gol

d PS

U L

osse

s 140 90% 120

88% 100

86% 80

84% 60

500W PSU 82% 40 1200W PSU Gold Eff 80% 20

0 78% 0 200 400 600 800 1000 1200 1400

Output Power 4 proc system operating range

dual proc systemoperating range

Effic

ienc

y

94%0.8 1

% LoadGold PSU Losses Which net loss0 0.2 0.4 0.6

requirement do we use?180

160 92%

Max

Gol

d PS

U L

osse

s 140 90% Option #1120

88%100

86% Option #28084%

60500W PSU 82% 40 1200W PSU Gold Eff 80% 20

0 78%

0 200 400 600 800 1000 1200 1400

Output Power4 proc system operating range

dual proc systemoperating range

Effic

ienc

y

Gold PSU Losses 0 0.2 0.4 0.6 0.8 1

% Load

180 94%

160 Option #1 92%

Max

Gol

d PS

U L

osse

s 140 90% 120 Option #2

88% #1 not as good for100 high power PSUs

86% 80

#1 not as good for 84% low power PSUs60 500W PSU 82% 40 1200W PSU Gold Eff 80% 20

0 78% 0 200 400 600 800 1000 1200 1400

Output Power 4 proc system operating range

dual proc systemoperating range

Effic

ienc

y

Gold PSU Losses 0 0.2 0.4 0.6 0.8 1

% Load

180 94%

160 Option #1 92%

Max

Gol

d PS

U L

osse

s 140 90% 120 Option #2

88% 100 #2 not capable

using low power PSUs 86% 80

84% 60 #2 not capable

500W PSU using high power PSUs 82% 40 1200W PSU Gold Eff 80% 20

0 78% 0 200 400 600 800 1000 1200 1400

Output Power 4 proc system operating range

dual proc system

Effic

ienc

y

operating range

0 10 20 30 40 50 60 70 80 90

100

PS

U lo

sses

(wat

ts)

570 1100 650 460 750

Curve Fitting with NPL Single O/P Gold PSUs (best in rating)

0 200 400 600 800 1000 1200

PSU Load (watts)

Proposed PSU Requirements for Storage Spec • Create three PSU categories for ENERGY STAR specifications

• Single O/P PSU category • Same efficiency and PF requirements as CSCI and 80+ Gold requirement for single O/P PSU

• Measurement excludes fan power • Multi O/P PSU category

• Same efficiency and PF requirements as CSCI and 80+ Silver requirements for multi O/P PSU

• Measurement includes fan power • Multi O/P PSU category for storage

• Efficiency and PF requirements should be based on SNIA’s Green Storage TWG – expected late Oct’2009

• Special consideration that data center storage PSU’s max operating load is 50% of its rated power load.

• Measurement excludes fan power • The storage power supply fan is used to cool the entire storage chassis, not

just the power supply. • Efficiency testing conducted at 230V for all power supplies

• Maintain PSU consistency and alignment for all ENERGY STAR specifications

Proposed PSU Requirements (cont) • Withdraw consideration of system level PSU

requirement – consideration from Tier 2 server • Idle load power specification encourage right sizing

•The definition of idle and storage idle power requirements will need to be refined relative to server requirements

• Use adders to address redundant power supplies • Estimated system NPL based on efficiency curves for

installed power supply • Focus on system level requirements rather than additional

PSU requirements • CSCI supports SNIA and other industry efforts to

develop storage performance/efficiency metrics butrecognizes the complexity of such a metric

•ENERGY STAR needs to move cautiously when defining storage performance and energy metrics given the complexity ofenterprise storage systems.

Proposed Accuracy Requirements for Storage • Accuracy requirement of ±5% accuracy with a

maximum error of ±10W per PSU• As the load decreases the ability to accurately measure

the power becomes increasingly difficult Æ maintain max error of ±10W for loads less than 100W.

• Requirement should be per PSU • Fixed maximum system error becomes increasingly

difficult for systems with redundant and additional PSUs • Eliminate specialized power metering solutions

• Sacrifices additional power losses to accurately measure input power

• Unnecessarily drive up PSU costs

Other questions: www.climatesaverscomputing.orgkathleen.m.fiehrer@intel.com