Impact of IT Trends On

High Availability
Power Architectures
Arunangshu Chattopadhyay
Director, Power Products Marketing and Head of Technical Support
Vertiv Asia

1
Agenda
▪ IT Trends and Their Impact: Driving the Evolution of the Gen 4 Data
Center

▪ Disruptors
▪ Electrical Designs and Architecture Evolution

▪ MTDC Trend and Stranded Capacity

▪ Rapid Deployment Need

2
Commercial In Confidence
Emerging IT Trends
▪ Devices everywhere. The
“connected devices” model
is becoming real and
expansive.

▪ The Time Factor. More

applications are demanding
near real-time response.

▪ Organizations are re-

evaluating how to best
deliver needs from
“core to the edge”.
IT deployments evolve to meet those demands
Deployment Scenario IT Needs

We are witnessing Edge 2 Core Capacity + deployment

a new IT Near Real-Time speed from edge to core.
Scalability. Economical.
deployment cycle, The 4th Generation
we call
Edge2Core Cloud Quicker large-scale
deployment.
(Public / Private)
to meet the new Greater efficiency.
demands of a
massively and Distributed IT Enhanced availability
rapidly connected Client-Server and flexibility.

society.
Provide power
Mainframe and cooling.
Edge2Core “Must Haves” + Never Go Down
• High Efficiency
• Low 1st Cost ‒ Certified efficiency and capacity
‒ Buy only what you need ‒ Low peak power utilization
‒ Buy infrastructure when you sell capacity ‒ Low air leakage
• High Asset Utilization ‒ High unit efficiency
‒ Eliminate stranded capacity – space, power, cooling ‒ Adaptive architectures and technologies
‒ Architecture that allows shared capacity across multiple systems (at the PDU) • Low Maintenance + Ease of Operation
• Rapid Scalability ‒ Remove system complexity, be consistent across sites
‒ Lowest installation cost
Save Me Money ‒ Rated for severe environments
‒ Grow while keeping infrastructure operating
Make Me Money ‒ Easy serviceability
‒ Optimized supply chain management
Lower My Risks ‒ Modular, replaceable components
‒ Add as capacity is sold
Make It Easier to Manage • Automated Control and Management
‒ Ability to scale in target increments – standardized building blocks or systems ‒ Remove complexity and human error
• High Reliability ‒ Be consistent across sites/locations
‒ Fault tolerant, ISO architectures ‒ Scale faster
‒ Component redundancy Best Practices At Each Location • Advanced supply chain management
‒ Compartmentalized -- Including systems, designs, partners, parts & labor

Local/Regional Data
Core / Central Data Centers Operations Edge/IT Spaces
Integrated Solutions/Services Integrated Solutions/Services Integrated Solutions/Services

Power Thermal Power Thermal Power Thermal

Monitor/ Control/Management Monitor/ Control/Management Monitor/ Control/Management

Commercial In Confidence 5
Impact of disruptive technologies

Source: Tony Seba

Disruptive Strategies
• AC UPS architecture shifts
‒ 2N+1 moving to 2N, 2N moving to reserve
Debate on which reserve architecture is the best fit
‒ Simpler configurations, central moving to distributed static switch
• Move to higher distribution voltages (both AC and DC)
• Decentralization of backup into the row and even into the rack
• Introduction of Lithium Ion Batteries into the critical space energy storage
landscape
• Hyper scale users are moving the needle with their scale
‒ Building their own data centers and experimenting with Pre-Fabricated Data Center
Deployments
‒ “Tuning” server HW and infrastructure to meet specific IT needs
• Emergence of Smart Power Applications
 Edge2Core: Power Management Best Practices Power

Modern Power OPEX CAPEX Space

Benefits
Best Practices Savings Savings Savings
UP UP UP
Flexible power location
S* S S
Deploy Emerging Increase utilization rates
Architectures Configure for SLAs
Deploy quicker

Choose high power Maximize kW footprint

density UPS Redeploy for more IT

Reduce battery space

Pilot Lithium Ion
Lower replacement cycles
energy storage Greater location flexibility

Add power as needed

Leverage modular,
Reduce 1st costs
scalable power Deploy quicker
Enhance PDU flexibility
Use flexible, ISO,
Maintain protection
high capacity PDUs Optimize space 8

Commercial In Confidence
Uptime Tier Classification Basics

• Tier I – Basic Capacity • Tier III – Concurrently Maintainable (Break One OR Fix One)
‒ Site-wide shutdowns are required for maintenance or repair ‒ Each and every capacity component and distribution of the path in a
‒ Capacity or distribution failures will impact the site site can be removed on a planned basis for maintenance or
replacement without impacting operations
• Tier II – Redundant Components ‒ The site is still exposed to equipment failure or operator error
‒ Site-wide shutdowns for maintenance are still required ‒ The preference is that all IT equipment is dual powered
‒ Capacity failures may impact the site
‒ Distribution failures will impact the site • Tier IV – Fault Tolerant (Break One AND Fix One)
‒ An individual equipment failure or distribution path interruption will not
impact operations
‒ A Fault Tolerant site is also Concurrently Maintainable
‒ Requires autonomous response to failure and notifies actions taken
Tier 3, Concurrent Maintenance
UPS System w/Redundancy and Multiple Power Delivery Paths for Concurrent Maintenance

IT Load
Tier 4, Fault Tolerance
Fully Redundant UPS System with Redundant Power Delivery Paths for Fault Tolerance

IT Load
Source: The Uptime Institute

Tier I Tier 2 Tier 3 Tier 4

Availability Experience (typical) 99.67% 99.75% 99.98% 99.99%
Typical 2(N+1)/2N Deployment
Utility Utility Performance:
• Concurrent Maintenance
Service Entrance Gen Gen Gen Service Entrance
• Every system/component can
be taken offline
Building Mechanical Mechanical Building
non-Critical non-Critical • Fault tolerant UPS, mechanical and
UPS In ATS
ATS ATS
ATS UPS In
downstream distribution

U U U U
Cooling
Cooling Cooling
Cooling Typical Characteristics:
P P P P
Equipment
Equipment Equipment
Equipment
S S S S • 2N Utility, 2(N+1)/2N UPS and 2N
UPS Out UPS Out
Distribution
P A A P

STS
STSs STS
STSs
• Redundant Generator(s)
PDU
PDUs CRAC
CRACs Racks CRAC
CRACs PDU
PDUs • N+x Cooling
R R
RDC
a Servers
c Servers
a RDC • ATS and STS used for enhanced
c
k k reliability and improved maintenance
Servers
P P • IT equip dual powered
D Servers D
U U
Reserve Architecture Improves Utilization and Efficiency

UPS UPS* UPS UPS* UPS* UPS UPS

2 x 2N Dual Bus Data Center Reserve Bus Data Center

~50% Utilization Rate ~75% Utilization Rate

• Enterprise traditionally utilizes high availability dual bus architecture

• Business models that demand higher utilization rates and can accept
lower reliability are turning to reserve (“catcher”) architectures
• Helps you move up the efficiency curve
Block Redundant Reserve System
Benefits
• Lower capex
• Smaller footprint
• Higher utilization
(above 50%) and
efficiencies
• Higher flexibility &
scalability

Disadvantages
• Added risk to loads
• Loads share reserve
• Manage transfer for
safe maintenance
Tip: Ensure reserve capacity is
not overloaded.
Distributed Reserve System

Benefits Disadvantages
• Lower risk using interleaved STS • Requires accurate load knowledge and management
• Less complex switchgear required for safe transfer
• Reserve redundancy achieved via unused UPS
capacity
Example of DYNAMIC BLOCK REDUNDANT System
System #1 System #2 Reserve System
Gen Gen Gen
Utility #1 2.5MW Utility #2 2.5MW Utility R 2.5MW

400v 480v 400v

1000kVA/kW

1000kVA/kW
1000kVA/kW

1000kVA/kW

1000kVA/kW

1000kVA/kW
ATS 600A ATS 600A

UPS

UPS
UPS

UPS

UPS

UPS
400v Mech 400v Mech

CPMS
1200A 1200A MIB MBB 2500A

Global Service
1200A 400V 1200A 400V 1200A 400V 1200A 400V

AC Power
3x800A 3x800A 3x800A 3x800A LBB/2500A 2500A

2500A Bus
A B C A B C A B C A B C
2500A LBB Bus
STS 600A STS 600A STS 600A STS 600A STS 600A STS 600A STS 600A STS 600A STS 600A STS 600A STS 600A STS 600A

500kW 500kW 500kW 500kW 500kW 600kW

500kW 500kW 500kW 500kW 500kW 500kW 600kW
500kW

400V 2500A 400V 2500A

IT A-B IT B-C IT A-B IT B-C IT A-B IT B-C IT A-B IT B-C 6x800A 6x800A

Trellis
IT A-C IT A-C IT A-C IT A-C

Customer IT

x 12
Space
x1
Maximize IT kW: Optimize Gen, ATS, UPS, Distribution
16
 Summary of MTBF Results
Summary of MTBF results with gensets Summary of MTBF results with gensets
Voltage failing at A and B simultaneously Voltage failing at A and B simultaneously

System MTBF MTTR Availability Relative System MTBF MTTR Availability Relative

hours years hours % MTBF hours years hours % MTBF

2(N+1) 9.01E+06 1,028 3.09 99.99997% 100% 2(N+1) 9.01E+06 1,028 3.09 99.99997% 100%
2N 6.98E+06 797 4.43 99.99994% 77% 2N 6.98E+06 797 4.43 99.99994% 77%
Distributed redundant with N UPS in reserve Distributed redundant with N+1 UPS in block reserve
3-to-2 6.23E+06 712 4.64 99.99993% 69% 4-to-2 7.38E+06 842 4.37 99.99994% 82%
4-to-3 5.41E+06 618 4.78 99.99991% 60% 5-to-3 6.78E+06 774 4.58 99.99993% 75%
5-to-4 4.78E+06 546 4.86 99.99990% 53% 6-to-4 6.27E+06 716 4.70 99.99993% 70%
6-to-5 4.28E+06 489 4.91 99.99989% 48% 7-to-5 5.83E+06 666 4.79 99.99992% 65%

7-to-6 3.88E+06 443 4.95 99.99987% 43% 8-to-6 5.45E+06 622 4.85 99.99991% 61%

8-to-7 3.54E+06 405 4.97 99.99986% 39% 9-to-7 5.12E+06 584 4.90 99.99990% 57%

9-to-8 3.26E+06 373 4.99 99.99985% 36% 10-to-8 4.82E+06 551 4.93 99.99990% 54%

10-to-9 3.02E+06 345 5.01 99.99983% 34%

High availability can be achieved – but be prepared for more failures!

Commercial In Confidence
The MTDC World

Commercial In Confidence18
Biggest Wholesale Data Center Deals – 2018*
1. Facebook: 72MW from CloudHQ in Manassas, VA
Multi-Tenant Data Center
2. Salesforce: 26MW from QTS in Manassas, VA (MTDC) leasing more than
3. Microsoft: 25MW from CloudHQ in Manassas, VA doubled in 2018 compared to
4. Facebook: 25MW from DLR in Ashburn, VA 2017
5. Microsoft: 20MW form CONE in Ashburn, VA Leasing activity in Northern
6. TBD: 20MW from DLR in Ashburn, VA Virginia (NOVA) for data centers
7. Streamcast: 15MW from Switch in Las Vegas, NV was more than two and a half
8. HUG (PBB): 15MW from Skybox in Houston, TX** times all other markets combined
in 2018
9. Microsoft: 14MW from RagingWire in Ashburn, VA
10. INAP (BoA): 10MW from Lincoln in Chandler, AZ Overall rental rates have
decreased as a result of new
operators and investors and
hyperscale users demanding
volume pricing
*Source: North America Data Centers **Immersion cooled HPC

Commercial In Confidence19
Strategies for Eliminating Stranded Capacity
• Optimize UPS ratings
Today some operators choose sub-optimal UPS ratings: 750kW vs 800kW, 1000/1100kW vs 1200kW. This is commonly
done for procurement tension vs capacity optimization

• Drive to 100% red line utilization

While red lines have moved up recently they are still typically less than 100% (90 to 95% UPS and battery). This could be
due to a number of factors: inconsistent full load capacity from vendors, lack of real time power consumption data, and
reluctance to oversubscribe

• Segregate load availability demands and oversubscribe

Today most operators target 5 x 9’s to one cord. If the loads could be segregated into 5 x 9’s and 3 x 9’s and load
distribution control and management were deployed overload capacity could be leveraged

• Reduce battery run times to only what is needed to bridge to the generator and run at higher temps
Battery run times have been coming down steadily but have stalled in the 5 min range at 25C. Some of this is due to
technology limitations: for example VRLA mfrs will not warrant their batteries for run times less than 5 min at 25C, LIB
and TPPL allow for 2 min at 30C. Operations teams that have repeatable transfer times can target sub 2 minutes of
battery run time
Commercial In Confidence20
Strategies for Eliminating Stranded Capacity
• Increase block reserve ratios
• Today most block reserve systems stop at 5+1 or 6+1
• Availability projections show 5 x 9’s operators can double that ratio and still meet 5 x 9’s
• Additional capacity can be picked up by converting reserve UPS to IT

• Value engineer system and power distribution gear

• Using coupled maintenance bypass gear vs split input and output gear saves bkrs and gear sections
• Single input vs dual input UPS
• Simplification of UPS paralleling tie
• Integrated critical output distribution (COD) vs standalone COD boards
• Fewer larger direct coupled PDUs vs multiple smaller distributed PDUs
• Elimination of redundant PDUs for ease of maintenance (it is a 100M+ hr MTBF piece of equipment)

All of the above depend on using the reserve for one of its primary roles, as a reserve during maintenance

Commercial In Confidence21
Normalizing in data centers?

Designs Equipment & Facilities

22
What Experts are thinking?

SIZE SCALE SPEED

23
Pre-fabricated data center value proposition
• Modularity and Scalability Modules & Containers
• Customer Need (Up to 3MW)

‒ Incrementally Build Capacity to Match Demand

‒ Or Build BIG Now!
• Our Differentiation: Innovative Designs
• “Data Center as a Solution”
‒ Fully-integrated systems
‒ Broad expertise enables optimization at the building block and data center levels

Buildings
(1MW and higher)

24
1.6MW Skid One line and layout

25
Distributed Power Infrastructure Converged Power Infrastructure

© 2019 Vertiv All Rights Reserved 26

Liebert APT
Adaptive and Insightful Power-train

© 2019 Vertiv All Rights Reserved 27

Summary
• IT demands at the edge are reverberating back to Delivering a Gen 4
the core, creating challenges
• Witnessing the shift to a new Generation of IT
data center requires
operations to: the holistic and
▪ Enable the processing of critical data in near real- harmonious
time across the entire IT landscape. integration
▪ Meet demands for growing capacities.
of the edge back to
▪ Deliver deployments quickly.
▪ Utilize building blocks that scale appropriately. the
▪ Make it economical. core facilities
• Improve Utilization and balance it against
Availability

28
Commercial In Confidence
VERTIVTM ENABLES THE
VITAL APPLICATIONS OF
OUR DIGITAL WORLD
DATA CENTERS
Micro, Edge, Cloud, Enterprise,
Prefab, Colo, Hyperscale

COMMUNICATION
NETWORKS
Small Cell, Macro Sites,
Central Office, Data center

COMMERCIAL AND
INDUSTRIAL
Healthcare, Rail, Transportation,
Power Generation, Oil and Gas