UPS For Efficient Data Centers

Eaton Corporation
 Agenda
(1) Introduction – Basics About UPS’s
(2) Trends and Challenges Affecting The Power
Infrastructure Of Your Data Center
(3) Selecting The Right Power Infrastructure For
Your Data Center
(4) The Quest For Efficiencies in Data Centers:
Addressing the Reliability vs. Efficiency
Paradigm of UPS’s
(1) Introduction – Basics About UPS’s
 What is a UPS?
• UPS stands for Uninterruptible Power System

• Acting as an interface between the mains and sensitive applications,

UPS’s supply the load with continuous, high quality electrical power
regardless
dl off the
th status
t t off the
th mains.
i

• UPS’s deliver a dependable supply voltage free from all mains

disturbances, within tolerances compatible with the requirements of
sensitive electronic devices.

• UPS can also provide this dependable voltage independently by

means of a power source (battery) which is generally sufficient to
ensure the safety of individuals and the installation.
 Why do we need a UPS?
• Digital equipment (computers, telecom systems, instruments, etc.)
use microprocessors that operate at frequencies of several mega or
even giga Hertz, i.e. they carry out millions or even billions of
operations per second.

• A disturbance in the electrical supply lasting just a few milliseconds

can affect thousands or millions of basic operations

• The result may be malfunctions and loss of data with dangerous

(e.g. airports, hospitals) or costly consequences (e.g. loss of
production).

• That is why many loads, called sensitive or critical loads, require a

supply that is protected against distribution system disturbances.
The image cannot be display ed. Your computer may not hav e enough memory to open the image, or the image may hav e been corrupted. Restart y our computer, and then open the file again. If the red x still appears, y ou may
hav e to delete the image and then insert it again.
Why do we need a UPS?

Source: CEMEP UPS European Guide

 How does a UPS work?
• UPS are generally made up of three main sub-
assemblies
– a rectifier-charger to transform the alternating current into direct
current and charge the battery

– a set of batteries (generally lead-acid type) enabling energy to

be stored and instantly recovered as required over a 5 to 30
minutes period
period, or even more

– an inverter to convert this direct voltage into an alternating

voltage that is perfectly regulated and filtered in terms of voltage
and/or frequency
 How does a UPS work?

Line
Interactive

On Line
Double
Conversion
 The 3 UPS Topologies
• Passive stand-by ("Off-Line")
– Basic UPS for applications for slightly disturbed environments
 E.g. individual PCs & their peripherals, home entertainment

• Line interactive
– Standard & most efficient UPS for IT applications
– For slightly disturbed environments in terms of frequency, but
accepting voltage fluctuations
 networked PCs,
PCs Servers
 network peripherals (printers, external disks, routers, etc.)

• Double conversion ("On-Line")

( )
– For disturbed environments, in terms of voltage, frequency,
harmonics, micro-outages and outages
 Critical servers & Datacenters
 Telecomm nications
Telecommunications
 Medical applications (scanners, IMR, etc.)
 Disturbed industrial environments, emergency lighting, etc.
 3-5-9 Concept

1. Power Failure 6. Switching Transient

3-5-9 concept 2. Power Sag 7. Line Noise

3  Off-Line
5  Line Interactive 3
3. Power Surge 8. Frequency Variation
9  On-Line Double
Conversion
4. Undervoltage 9. Harmonic Distortion

5. Overvoltage 5
9
(2) T
Trends
d andd Ch
Challenges
ll Affecting
Aff ti The
Th
Power Infrastructure Of Your Data Center
Past Data Center Power Deployment
 Why Is Power More Important Today?
Protecting Protecting Data Protecting
H d
Hardware L
Loss and
d C ti it off
Continuity
Equipment Corruption Service

1940 1950 1960 1970 1980 1990 2000 2010

Industrial Digital Economy Services

Economy Economy
 The Power of IT
• First g
generation p
power thinking
g

If the power Power is a

I need a
ain’tt there,
ain binary state:
UPS I need
UPS, d
you ain’t it’s either
a generator.
there. ON or OFF.

Power presence
 Real Power
• Just because y
you have power,
p doesn’t
mean you’ve got good power
– It’s onlyy the beginning!
g g
• Power is not a binary thing

Difference bet
between
een reacting to a
catastrophe and preventing a catastrophe
 The Power of IT
• Second g
generation p
power thinking
g

How
good is How
How good What much
is the the un- damageg
I will have energy
power conditioned conditio can I
power ned prevent if I am I
behind my power in knew what
UPS? to look for? consumi
front of ng?
my UPS?
Power quality
You Monitor CPU,
CPU Storage…
Storage
 Incremental Evolution

Power is a How good is How good is How much

If the power I need a I will have What damage
binary the the un- energy
ain’t there, UPS, I power can I prevent if
state: it’s conditioned conditioned
you ain’t need a I knew what to am I
either ON power behind power in front
there. generator. look for? g
consuming?
or OFF.
OFF my UPS? of my UPS?

• To drive additional 9’s into the reliability equation

– More and more detailed visibility will be required
– With more and more expertise required
Today’s
Today s Data Center Infrastructure
 Market Trends
• Energy efficiency
– E
Energy usage ini d
data
t centert has
h grown 2 – 3x3 iin th
the llastt th
three
years
– Through 2009, energy costs have emerged as the second-
highest operating cost (behind labor) in 70% of DC facilities
worldwide*
– Corporate responsibility to reduce greenhouse gas emissions is
on the rise

• Flexibility / scalability
– Moves
Moves, adds and changes (MAC) are the norm in any data
center, where change is the only constant
– Ability to scale support infrastructure with growth of IT equipment
is increasingg

* Source: Gartner; Meeting the DC power and cooling challenge

 Technology Advances
• High density blade servers and rack mounted storage arrays
– Being promoted as the saviors of the modern day space constrained
data center manager

• Virtualization of IT applications
pp
– Can increase server utilization from an average of 15% to over 80%

• Virtualization of disc storage

g
– Can allow much higher utilization of the attached storage, therefore
reducing the number of discs required

• Data center consolidation

– Can reduce operations costs, freeing up capital for additional IT
technology
 Technology Challenges
• Increased power density
– M
Mostt traditional
t diti lddata
t centers
t constructed
t t d 3 tto 10 years ago were
engineered to accommodate 3-5 kW/rack; the new technologies
can pack in IT equipment with a power and heat load as high as
30 kW p per rack.

• Dynamic virtualization environments

– C
Can fforce lloads
d around
d diff
differentt areas off th
the d
data
t center,
t
stressing the environment even further

• Dealing with existing infrastructure

– Older power (UPS system) and cooling system products are
much less efficient than newer systems - contributing to the
problem
bl
 Power On…
On
• You require more efficient power and cooling systems, reducing
operational costs,
costs and allowing you to re-invest
re invest savings in improving
your business
• You need more flexible systems to allow you to more easily change
your IT environment to meet the business needs
• You need higher reliability and availability – It’s not acceptable to be
off line as the business requires the systems to be on-line to survive
– Power
P & cooling
li systems
t should
h ld bbe redundant
d d t & expandable
d bl tto N+x
N+
• Interoperability of all equipment is a must, ensuring seamless
deployment and operation of new systems
• Physical space on the data center floor or in the building is very
expensive - so power and cooling systems must be highly efficient in
size… as well as performance
 (3) S
Selecting
l ti TheTh Right
Ri ht Power
P
Infrastructure For Your Data Center
 >100
Racks

1 to 5
Racks
Centers

Racks
Racks

Wiring
5 to 20
Centers

Centers
20 to 100

Closets /
Large Data

Small Data

Server Room
Mid Size Data

1 - 10kVA UPS

10 – 60kVA UPS

60 - 200 kVA UPS

S

> 200 kVA U

UPS
Distributed
d UPS
Zone UPS
Z
Matters

Centralized U
UPS
Size Matters…

In the R
Rack
Insiide Computer Room

Outside C
Computer Room
m

Rack Power Strip (ePDU)

Floor// Rack Power D

Distribution
 Size Matters…
Matters Centralized UPS
Centralized Computer Room
–Scalable for buildouts
–Ongoing
O i maintenance
i t iis simplified
i lifi d
UPS

Zone UPS
Computer Room
Zone UPS
–Scales with computer room growth
–Maintenance more intensive than UPS
centralized
t li d UPS

Rack Mounted
Computer Room
Rack
R kMMounted
t d UP
S UP
–Provides ultimate scalability UP S UP
S UP
–Limits impact of UPS failure to an S UP
S UP S UP
enclosure(s) S UP S
S UP
–For
F high
hi h d
density
it areas off your d
data
t S
center
Many Other Factors To Consider
• A range of factors must be taken into account when selecting the
right power infrastructure for your data center
– Number of racks
 Small data center, computer room, network closet: (<20 racks)
 Medium data center: (20 – 100 racks)
 Large data center: (100+ racks)
– Power density per rack
 Low density (1-4 kW/rack)
 Medium density (5 – 9 kW/rack)
 High density (10-15 kW/rack)
 Ultra High density (>15kW/rack)
– Data center availability (Uptime Institute)
 Tier I (basic)
 Tier II (redundant N+1 components)
 Tier III (concurrently maintainable)
 Tier IV ((fault tolerant))

Every data center is unique based on business needs

 A Typical Large Data Center
Bus “A” Bus “B”

Incoming Service Gear UPS

Generators

Server racks

Substation Transformers Fire Suppression

Cooling

Operations Center

Security Access
 Example Of Large Data Center Project
• Sentrum Data Center – Woking, UK

Chillers

UPS

Data Suite

Generators
 In A Nutshell
• There are a lot of technology changes going on that can
effect
ff t today’s
t d ’ data
d t center
t design
d i
• The only thing common in two Datacenters is the
increasing complexity and number of components
• Power remains a critical component
• Monitoring, Data, and Management puts you in Control
• And don’t forget Services
(4) The Quest For Efficiencies in Data Centers:
Addressing the Reliability vs. Efficiency
Paradigm of UPS’s
 Still Growing Demand
• Economic downturn has significantly impacted the data centre market
– Postponed investments and projects in many companies
 However, projects haven’t been cancelled
 Many of them are being released this year

• Market drivers for power-hungry

power hungry society still valid today
– Growing demand for highly available and reliable power in our “online everything”
societies
– IT is not limited to businesses anymore
 Intrinsic
I t i i tot every aspectt off our society
i t

• Increasing demand for highly efficient and reliable power in data centers
– Availability and its targeted "five
five 9
9’s"
s remain the number one requirement when
selecting the appropriate UPS
– Increasing need for power and increasing cost of power
 Energy efficiency shifting from "nice-to-have" to "need-to-have“
 Small % of efficiency
y brings
g tremendous savings g
 Availability Remains At The Top
• Uptime and quest for five 9’s remain a key focus for data center
managers
– When selecting UPS’s for data centers, double-conversion has been the
reference to achieve highest level of reliability

1. Power Failure 6. Switching Transient

2 Power Sag
2. 7 Line Noise
7.
3-5-9 concept
3  Off-Line 3. Power Surge 3 8. Frequency Variation
5  Line Interactive
9  On-Line Double
Conversion 4. Undervoltage 9. Harmonic Distortion

5. Overvoltage 5
9
 Energy Efficiency from nice
nice-to-have
to have to need
need-to-have
to have

• Energy costs and regulatory compliance will continue to place

organizations
g at risk
– Social, economic and regulatory pressures to reduce carbon emissions and
energy costs globally

– Increasing Energy Costs

 Through 2009, Energy costs will emerge as the second highest operating cost (behind
labor) in 70% of the Data Center facilities worldwide (source: Gartner)
 Cost to run a data center > than cost to finance it initially

– Sustainability
 80% of CEO’s view sustainability as impacting brand value (source: 2008, McKinsey)
 31% say they want to reduce their environmental impact (source: 2008, McKinsey)
 IT accounts for 2% of the global CO2 emissions, as much as the airline industry
(source: 2008, IBM, UK)

– Regulations
 82% of executives expect some form of climate change regulation within 5 years
(source: 2008,
2008 McKinsey)
 European Union Code of Conduct for Data Centers
 UK Carbon Reduction Commitment (2010)
 Addressing Reliability vs
vs. Efficiency Paradigm
• The key challenge for Data Center managers
– How to achieve higher efficiencies without compromising reliability
 Historical trade-off
 Double-conversion: maximum level of protection at a cost
 Other legacy modes: compromise on reliability

• The solution = two new game-changing UPS technologies

– Variable Module Management System (VMMS)
 Maximized efficiency in double conversion mode
 Both in single- and multi-UPS systems applications
– Energy Saver System (ESS)
 The ultimate savings: 99% efficiency
 Automatic fast transfer to double conversion mode
when needed (in less than 2 ms)

 Higher
g System
y Efficiency
y
 No Compromise On Data Center’s Reliability
Variable Module Management System
• In double conversion mode, the efficiency of any UPS varies
F
depending on the % of load
A
C – Highest efficiency when close to full capacity
T • UPS systems rarely loaded at full capacity
S – This
Thi iis th
the ffactt iin redundant
d d t systems
t as used
d iin mostt d
data
t centers
t

C
H
A
L
L
E
N
G
E

 How to maximize efficiency potential of UPS systems with

lighter loads
 Variable Module Management System
• How to maximize efficiency with lighter loads in double conversion
mode
– Especially in multi-UPS and redundant system configurations

• Solution #1 (basic)
( )
– Concentrate load on certain UPS’s to maximize UPS load level and
overall system efficiency
 Some energy savings
 Limited to multiple-UPS
multiple UPS systems (with several UPS in parallel)
 Still not optimal

• Solution #2 (more advanced & efficient)

– Automatically optimize efficiency at UPM level  VMMS
 Concentrate the load on certain UPM’s to maximize
overall system efficiency
– Onlyy p
possible on modular UPS’s
 Not limited to multiple-UPS systems
 Achieve even higher optimization thanks to UPM’s modularity
Variable Module Management System
• VMMS Principle
– Example With Same Load Applied To Different Multi-UPS
Multi UPS
Configurations
 Variable Module Management System

• VMMS Principle
– Maximizes % load of each UPM
 Optimizing overall system efficiency
100%

System Efficiency gain

thanks modular UPS design
with VMMS capability
90%

80%

20% 40% 60% 80% 100%

% load of each UPM in Case1 when

% load of each active UPM in
using UPS without VMMS
Case3 when using VMMS
 Variable Module Management System
• VMMS Benefits
– Maximized efficiency in double conversion mode  higher
efficiency

• Typical applications where VMMS is particularly efficient

– Redundant N+1 and 2N systems UPS’s
 Li
Lightly
htl loaded:
l d d typically
t i ll operate
t att lloads
d < 45% lload
d llevell where
h
efficiency is not optimal
– Data Centers, especially when UPS system feeds dual corded
servers
– Any applications when load is not constant
 Energy Saver System
• Energy Saver System Principle
– Until recently
recently, upper limit for UPS energy efficiency in double-
conversion ~96%
 For 1 EUR spent on utility power, three to seven cents of it was used or
dissipated as heat by the UPS

– In developed countries, utility power is within industry specifications

most of the time
 Complete
p p
power conditioning
g is needed only
y rarely
y
 Yet traditional double conversion UPS’s continuously process utility power,
converting it from AC to DC and back to AC again
 Dissipating heat and wasting power at every stage.

– Energy Saver System technology delivers 99% efficiency by intelligently

adapting to utility power conditions and operating in energy-saving
mode most of the time
 Offering double
double-conversion
conversion protection
 when your load needs it
 As quickly as it needs it
 Energy Saver System
• Energy Saver System engages modules according to
input power
Input Power Quality
within
out
within
of rectifier
load tolerance
tolerance Active Modules

+10%

+ 10%
45Hz 55Hz
47Hz 53Hz
-10%

-15…-30%

at nominal voltage 400V/50Hz

 Superior detection and transition technology
 Energy Saver System
• Latest innovation in the area of UPS
– Providing
P idi the th highest
hi h t reliability
li bilit with
ith th
the hi
highest
h t efficiency
ffi i
 UPS power converter stages adapt operation to provide optimum
load availability and energy efficiency
 With advanced DSP (Digital Signal Processing) controls, power
conversion stages are “shifted into gear” only when needed

• What ESS does

 Maximizes efficiency when the input is within the acceptable
voltage and frequency window
 99% efficiency over wide load range - above 20% load

 Maximizes performance when the input is outside the window

 Full voltage and frequency independency available in just 2ms
 UPS seamlessly transitions through different operating modes as
needed
 only possible with transformer-free topologies
 Energy Saver System
• ESS Efficiency – 99% across the complete operating range
• 85% reduction in losses compared to legacy transformer
transformer-based
based UPS
• Continuous power tracking and proprietary DSP algorithms combined
with transformer free design topology ensures critical loads are always
p
protected
99% efficiency even at
100% lower load levels thanks to ESS

Double-conversion efficiency
m efficiency

of leading UPS’s

90%
System

Typical
Operations

80%

20% 40% 60% 80% 100%

% load
 Energy Saver System
• The bottom line

– ESS drives the UPS losses close to zero

– The energy savings from ESS typically recovers 100% of the
UPS cost over a 3 - 5 year period
– At 250kW of critical load, the savings is equivalent to 3 600 €
per year per point of efficiency gain

Calculation is based on 60% DCiE and CO2 emissions at 400g/kWh medium European level
 Energy Saver System
• Some form of multi-mode capability has been available
on UPS
UPS’s
s for years
– However, conventional multi-mode UPS’s have their limitations
 Trade-off between reliability and efficiency

• What differentiates ESS from traditional “ECO-Mode”

– Conventional ECO-mode is accomplished
p in one of two ways
y
 Power the load with Bypass Utility and wait for the UPS to start,
sync, and transfer to double conversion mode when disturbances
occur
 Long transition times to inverter
 Operate in line interactive mode with the inverter running for
faster transitions to double conversion mode
 Increased power losses resulting in reduced efficiency
– ESS solves
l both
b th issues
i
 It provides best possible efficiency, and instantaneous transfer to
double conversion when needed
 Closing thoughts
• Utility costs now account for 20 to 30 percent of data center
operating costs.
costs
– A million Euros worth of servers that was purchased in 2009 will
consume €1.2 million in electricity over a three-year operating life — a
figure
g that will only
y increase with rising y rates1
g utility

• The good news is that new, high-efficiency, multi-mode UPS’s can

dramatically reduce the costs and environmental impact of powering
the business
– More efficient allocation of power
– Reduced utilityy bills and total operating
p g cost
– Achieve more with available power and cooling systems — delaying the
point where those systems would have to be upgraded or replaced

Sources:
1. Underlying figures from The Invisible Crisis in the Data Center: The Economic Meltdown of Moore’s Law, Uptime Institute, 2007
Any questions?