Connectix Commercial Overview Dc Session 8 Using The Fear Model To Design Data Centres Compressed Pics
1. Data Centre Environments
CCS Commercial Overview Session 8
Using the FEAR Model in Data Centre Design
19th
November 2010
Paul Mathews MInstSMM, CCNA, MIEEE
Global Channel Manager
2. • Presented at the BICSI Winter Conference
Matt Parker RCDD, Stantec Consulting Services Inc
• Discuss how FOOTPRINT-ENERGY-ARRANGEMENT-REDUNDANCY helps the
management of data centre projects
• This presentation provides;
– Survey guide to apply basic design calculations to develop a flexible space plan and
gain a strategic position on the design team for DC projects
– Review of design and construction processes, linear Vs integrated design
– Use block design approaches to data centre
– Does not provide a single tool to use for designing a data centre as every facility is
different. Provides a knowledgeable background for common design processes
– Use best practice design as a sales technique for DC owners / architects
– Ideally useful for Tier I, Tier II – Tier II data centres
About FEAR Model
3. • The heartbeat of any business, designed to manage the flow, processing and
storage of information
• Must be reliable, secure and flexible to enable growth and reconfiguration
• A data centre can support small singular businesses through to thousands of
clients ecommerce facilities
• “A building or portion of a building whose primary function is to house a
computer room and its support areas,” according to TIA 942
• User centric-tool
• Combination of;
– Storage Area Network (SAN)
– High Performance Cluster (HPC)
– Enterprise Process Servers
Introduction to Data Centres
6. • ITS designers and consultants and have a larger impact on building design by
synergistic approaches with architects, owners and contractors
• Eliminates ‘pass it on’ processes;
– through sub-contracting; lack of project management and ownership of objectives
– Understand best practices for IT concepts and planning
– Control scope of works before bidding and contraction
• Increase the value of the IT designer / consultant
Integrated Design – what does it mean?
7. • Allows involvement with the facility owner;
– Addressing IT design during budget allocation
– Owner can make a value decision instead of budget decision
• Allows input into design processes;
– Developing peer relationships with traditional consultants / architects
– Ensure standards compliance for IT cabling
• Increase revenue for your company
Integrated Design – installer benefits
8. • F ootprint
• E nergy
• A rrangement
• R edundancy
About FEAR Model
9. • Define the facility for equipment quantity
• What type of process equipment is to be installed?
– Independent rack equipment
– Inter-dependent rack equipment
• Process support equipment?
– Test racks (burn-in, troubleshooting)
– IT / Network Connectivity (switches / routers)
• Non-process support equipment?
– UPS / PDUs (electrical distribution)
– HVAC / Mechanical Equipment
FOOTPRINT
10. FOOTPRINT
• All rack enclosures are relatively the same footprint – a block of 4ft x 2ft
(1200mm L x 600mmW )
• Agree on the size of racks for equipment
1200mm
600mm
11. FOOTPRINT
• Group racks together on multiple levels, to create modular planning cells
(this example uses a block cell of x 5 racks);
– Smaller data centres use a 5 rack configuration of modular cells to enhance
flexibility
1200mm
600mm
13. FOOTPRINT
• The paired cells become rows – where we
consider space allowances
• Aisles;
– Air movement for hot and cold zones
– Operator access and movement
• Non-process equipment electronics / M&E
management)
• These space allowances creating the core
planning blocks (discussed in more detail in
ARRANGEMENT)
14. • The energy required for a data centre is defined the same way as the
footprint, calculating the total kW needed for the process equipment
• What type of process equipment is to be installed?
– Independent rack equipment
– Inter-dependent rack equipment
– Today’s technology allows most equipment to be flexibility collocated
• Process support equipment?
– Test racks (burn-in, troubleshooting)
– IT / Network Connectivity (switches / routers)
• Non-process support equipment?
– UPS / PDUs (electrical distribution)
– HVAC / Mechanical Equipment
ENERGY
15. • Always calculate Energy PER RACK (Vs Energy per square
foot/metre)
• Allows energy distribution (cooling and energy) to move with
the FOOTPRINT
• Data Centre remains more adaptable to various processing
equipment
• ENERGY should be kept scalable and flexible;
– Not new concept for data centres
– Moore’s Law becomes evident (DC structures will change in 16-
18 months)
ENERGY
16. • AT DESIGN STAGE; Data Centre owners specify the equipment to be used in
the racks, advising a model number and product data sheet
• This can give Electrical Engineers only very basic information to be able to
calculate ENERGY requirements
• Data Centre infrastructure designers can use some baseline techniques using
general power requirements;
– Traditional rack equipment density uses 2 kW – 8 kW / rack
– High density processing equipment racks use 8 kW – 12 kW / rack
– 2010 beyond 12 kW / rack
• Additionally include the cooling needs;
– Input power requirements do not include cooling power energy
• Total Energy Formula to include distribution method, cooling density and
equivalent power needs
ENERGY
17. • The number 1 question for ENERGY in data centres is;
HOW MUCH POWERPOWER IS NEEDED?
• This can be calculated by adding the input power and power loss budget
together
• INPUT POWER BUDGET;
– Electricity require to operate the process equipment loads
• POWER LOSS BUDGET;
– Electricity required to remove heat (cooling equipment, air distribution equipment
etc)
Input Power + Power Loss = Total Power
ENERGY
18. ENERGY
• Calculations for capital and expenditure costing can only be approximated with the
manufacturers data sheet information
• Many designers can misinterpret or overstate power and cooling requirements
without correct data
• Data sheets need vital information including;
– Input power
– Heat dissipation
– Processing Equipment physical dimensions
19. ENERGY
• Data can be stated in different formats;
• Electricity and power supplies are the drivers for heat dissipation
and cooling
20. • The Power Loss value must consider inefficiency of using HVAC
equipment to remove heat;
– Cooling power generates accumulated losses from 2 level of heat
exchange;
• Chillers to condenser (for cooling equipment)
• Water to air (for air handling equipment)
• 3 Useful Rule of Thumb equations for planning;
– 12,000 BTU = 1 cooling ton = 350 Cubic Feet per Minute (CFM)
– 1 cooling ton = 1.2 kW electric power (consult ASHRAE guidelines)
ENERGY – Calculating Power Loss budget
21. ENERGY – further basic formulas for TOTAL POWER
• BTU = 1.08 * CFM (air flow) * ΔT (temp. rise)
• 1 BTU = 0.293 = 1 Watt = .001 kW = I (Current) x V (Voltage)
• An example to demonstrate equivalence from best data;
22. • Rack enclosure filled with 42 x 1u servers from Manufacturer A;
– Rack capacity = 42 RU, total servers = 42
– Average power data for 1U server = 400 W = 0.4 kW
– Heat dissipation = (400 W / 0.293) = 1365 BTUh
• Total electric power for a fully loaded average server rack is;
– (0.4 * 42) + (((1365 * 42) / 12,000) * 1.2) =
16.8 + (4.75 * 1.2) = 22.5 kW per rack
• Equivalent to a typical house 2500 sq ft into 8 sq ft of floor
ENERGY – 42U Rack Example
23. • Gives more flexibility, so power and cooling distribution can be
selected to fit the internal rack equipment
• kW / sq ft thinking leads to bulky, inflexible and incorrectly specified
process support equipment
• Modern day server racks typically utilise the following power
densities;
– Storage / Application Rack = 20,000 – 35,000 BTU, <8 kW / rack
– Network Racks = 5,000 – 10,000 BTU, 2 – 4 kW / rack
• Modern day server racks accommodate planned diversity (grouping
different systems together – e.g. not 10 HPC side-by-side, but adding
more cabling
ENERGY kW per rack Vs kW per sq ft
24. • Placing the process equipment around the building area with
consideration for;
– Architectural and Structural Design;
• New or existing building
• Adjacency and size of support spaces?
• Raised floor and finished ceiling?
– Non-Process Support Equipment;
• Power and cooling distribution methods?
• Entrance facilities and equipment spaces?
• Equipment size and quantities?
ARRANGEMENT
25. • The FEAR model is based on a new building concept, but can
be adaptable for an existing facility
• Design experience implies a raised floor AND finished ceiling
is the most flexible and efficient, creating hot and cold
plenums
• Suggested optimised height is 14 ft (floor to deck), with 18 –
24” for raised floor, 9-9.5” finished ceiling
ARRANGEMENT
26. • Does not specifically address support spaces such as chiller room, main
electrical room (should be consulted on a project by project basis).
• Assumes new building design has adequate mechanical, electrical and site
space
• Existing buildings can adapt the FEAR model to determine the max. power to
the building and make a model that will fit the available power (on a block
basis) to help owner observe if a facility is suitable for a design
• Provides new building conceptual cost plans, evaluation of sustainability or
suitability for existing spaces
• Non-process equipment – considers power and cooling distribution methods.
Raised floor plenums are not the most efficient for electrical distribution or
cabling due to cable tray, drop boxes, junction boxes within the plenum
distorting the HVAC system effectiveness
ARRANGEMENT – FEAR MODEL CONCEPT
27. • Power is from overhead feeds;
– Busway – allowing compact, flexible implementation (or
manufacturers power distribution panels)
– Conduit-and-box – familiar to electricians, lower first cost but
some issues with TCO (relocation)
• Network cabling routed overhead;
– Multi-level cable tray for copper or optical fibre
– Parallel wire baskets
ARRANGEMENT
28. • Primary Cooling distribution through FLOOR plenum;
– Proven system
– Cool air is delivered on equipment where needed
– Optimise rack efficiency using blanking panels
– FEAR model allows use of new point cooling methods (such as
spray directly on equipment)
ARRANGEMENT
29. • Entrance facilities and Equipment Spaces;
– Depend on other building issues
– Includes chillers, generators, centralised Ups
– By using the Integrated Building Process to define equipment
size and numbers, the FEAR model provides consultants with
plans for ERs when architects are laying out block plans
• Equipment size and quantities;
– Defined by PDUs and CRAC units
– Footprint of processing equipment
– Size of support equipment depends on average power density
per module
ARRANGEMENT
30. ARRANGEMENT (using FOOTPRINT model)
• 4 blocks of 10 racks
• Incorporate standard aisle spacing (2 rows
grouped into a block) ;
– 6ft between racks, providing high density
applications with air distribution tile on either
side of aisle to provide air and includes walk tile
in the middle – can reduce to 4ft if needed
• Hatched area provides space reserved for
process support equipment (CRAC, PDUs)
• Block dimension is 20ft x 50ft;
– Floor dimension is 25ft x 60ft, matching
common structural grid spacing of 25ft x 30ft (2
bays)
31. ARRANGEMENT (using FOOTPRINT model)
• CRACUs sized to rack power density;
– accommodate 4 CRACUs (for high density
applications)
– 2 power distribution units with redundant
capacity
• Within the racks;
– the back space houses the power distribution
as the hot aisle
– Front space faces the cold aisle where network
cabling is configured, equipment LEDs are
checked (as green for working) by network
technicians
32. ARRANGEMENT (using FOOTPRINT model)
• Blocks are modular – can be combined in full
or half to create any shape
• Blocks can be added until rack quantity
meets facility owners stated objectives
• Each block is self-contained and includes
support equipment
33. REDUNDANCY
• Must be specified as a tiering model by the facility owner in
advance
• Difficult to make Tier I/II into a III or IV at a later date
• Must consider process AND power AND cooling
• FEAR model covers hybrid solutions
34. REDUNDANCY
Tier Classifications (Uptime Institute)
Tier 1 Tier 2 Tier 3 Tier 4
Site availability 99.67% 99.74% 99.98% 99.99%
Downtime (hours per year) 28.8 22 1.6 0.8
Operations Centre n/a n/a required required
Active Capacity Components to support
the IT Load N N+1 N+1 N after any failure
Distribution Paths 1 1
1 Active and 1
Alternative
2 Simultaneously
Active
Concurrently Maintainable No No Yes Yes
Fault Tolerant No No No Yes
Compartmentalisation No No No Yes
Continuous Cooling
Load Density
Dependent
Load Density
Dependent
Load Density
Dependent Class A
35. • Tier 1;
– 99.671% availability (2 nines)
– Single path for power and cooling distribution
– No redundant components
• Tier 2;
– 99.741% availability (2 nines)
– Tier 1, add redundant capacity components (computer equipment)
• Tier 3;
– 99.982% availability (3 nines)
– Multiple power and cooling distribution paths
– Redundant components, one active path
– Concurrently maintainable
• Tier 4;
– 99.995% availability (4 nines)
– Same as Tier 3 with multiple active paths
– Fault Tolerant
REDUNDANCY
36. • FEAR model targets Tier I and Tier II facilities;
– A Tier IV data centre requirement is mostly specified outside of
the data centre footprint
– Tier III or Tier IV facility costs are non-linear (get expensive very,
very quickly) and require a strong business case
– Tier II classifications are the most common data centres;
• N+1 configuration
• Using FEAR model Tier II to Tier III can be scaled with a modest effort
(with early discussions with facility owner;
– Providing power density and scalability of Tier III without non-linear cost
increases
– Coordination of power and cooling equipment with building engineer
critical to achieving Tier III (provide the completed model to mechanical
designers)
REDUNDANCY
ITS Vendor selected when owner is moving in.
Works in commercial buildings but not in DCs
Linear Design – as the idea of a project moves from conceptual to actual, the change cost increases
Integrated design theory places ITS professionals (designers and consultants in the conceptual stages and moves from being at the construction end with the smallest sales value to working as design and consultancy stages (and increasing your financial value possibilities)
Process equipment – high density applications
Non-support process equipment –
Is the equipment independent or dependent
Example shows 3 different manufacturers of blade servers with over a 50% difference in energy requirements, which when installed into a modular data centre has a major implication for operating expenditure
80% of use of equipment can be power idle
Average power is the energy used when processing
Power supply max is often a guideline used by designers to ensure complete performance accountability
Taking the heat from the equipment (blowing cold air on processing equipment)
Discharge the heat somewhere – to either the air or refrigerant
BTU is the measurement for heat gain within a building (thermal dynamics)
CFM is air movement (useful as equipment manufacturers are now including information about how much air is needed to keep cool)
The DELTA flow and air flow determine the efficiency of data centre cooling systems.
2 pieces of raw data would be required to make a 5-10% estimation of ENERGY requirements
Support spaces being main chiller room, electrical room