Professional Documents
Culture Documents
Designs
Next Generation Data Center
August 8, 2007
Doug Lauterbach
Data Center Director, Bay Care Health System
Matt Schuster
Principal Consultant, BT INS
In today’s presentation…
• About BayCare
• Challenges that necessitated a new data center
• Planning the new facility
• A look at the new data center
• Your questions
About BayCare
• 9 hospitals in Tampa Bay area
• Largest full service community-based health system in
the region
• 17,000 team members
• 2,707 Beds
• 355,523 ER visits
• 6.3 million lab tests
Challenges that necessitated a new facility
Business needs driving IT
• In the process of a 7 year electronic medical record
project (moving towards paperless)
• Constant growth of digital imaging
• Increasing compliance regulations - JCAHO concerns
about information management and availability
• Increased clinical reliance on IT systems requires high
availability for quality patient care
– Pharmacy
– Lab
– Radiology
– Food and Nutrition
– Bed tracking
– Scheduling
Business Drivers of the new Facility
• Best Practices
• Business Continuity
• Business Expansion
• Capacity Planning or Performance Analysis
• Change in Application Architecture, Providers or Vendors
• Compliance
• Disaster Recovery
• Expense Management
• Increased Capacity
• Operational Optimization
• Regulatory
• Reliability
• Risk Management and Mitigation
• Technology Growth or Consolidation
• Technology Refresh
Existing facility challenge: space
• Short on space
– 8 separate spaces for IT equipment
– 4,700 sq. ft data center split into 4 spaces
• 30% of data center space used for circulation
• Large command center using a large portion of
raised floor space
• Building’s physical resiliency not up to hurricane
standards
• Design aspect to 40 watts/sf
• Shallow cabinets limit the ability to expand and distribute
power and low-voltage cabling
• Structure cabling unable to support:
– Blades
– VoIP
– Anticipated SAN growth
Existing facility challenge: power load
• Design didn’t support heat and power load
• Downtime related to power design and capacity
– circuit capacity lacking
– Distribution: not dual bus all the way to equipment
• Lack of available power threatening to limit new
technology adoption
• No standard power configurations
• No means or method to manage/measure power at the
equipment racks
Existing facility challenge: cooling
• Cooling was inefficient
– Not using hot-aisle/cold-aisle design
– Uneven temperatures through space
– Hot spots creeping up
• Equipment lifespan likely shortened due to heat
• 13-year old design, not optimized for blade configurations
• Not easily expandable
• Low 12” raised floor and under-floor low-voltage cabling
restricting air supply
• No available physical space for additional cooling
• Legacy CRACs were not communicating to efficiently cool
the space.
Existing facility challenge: visibility & control
• Monitoring – currently focuses on circuits that are tripped
(reactive), not designed for proactive data
• Temperature monitoring handled at the return air of the
precision air conditioner (not looking at hotspots)
• Little remote control for servers
• Lights-out management (iLO)
• Existing Liebert SiteScan equipment was not managing
non-Liebert equipment
• No temperature sensors at equipment location or around
the room
• No branch circuit monitoring
• 100% of lighting on 24x7
Existing facility challenge: managing change
• Provisioning is not standardized, automated or managed
• System deployments are based on:
– looking through racks
– personal knowledge
– gut feeling
– where uneducated users want placement
• Not managing capacity of:
– space
– circuits
– power
– weight
– heat loads
– network switch ports
Planning the New Facility
Objectives
• Provision for planned growth – addition of 100 servers per year
• Gain complete visibility and control and improve administration
• Improve system availability through full redundancy of the
power and cooling systems
• Improve fire detection and suppression
• Implement hot-aisle/cold-aisle design for maximum capacity
and energy efficiency
• Improve planning and provisioning
• Improve cable management
• Increased Mechanical and Electrical efficiencies
• Reduced operational costs
• Operate as a nearly lights-out facility
• Remote access into all systems
Building a Team
• Doug Lauterbach
Imaging Services Director (serving as Data Center Project
Director), BayCare Health System
• Chris Jenkins, Director Technology, BayCare Health
System
• Matt Schuster, Principal Consultant, BT INS
• Thomas Gooch, Local Liebert Representative, Emerson
Network Power
• Paul Carastro, Carastro & Assoc, MEP Engineers
• Mike Montecalvo, General Contractor, Solutions, Inc.
• Paul Schnitzlein, Architect, Harvard Jolly
Documenting Existing Equipment
• Document existing devices, power supplies and port
counts per device at the rack level
• Determine/Graph device averages per rack to determine
cabinet profiles
• Each piece of equipment assigned a number and
quantities and loads were considered. Classifications
included:
– Legacy
– Low
– Medium
– High
– Blade
– IBM RISC
– Network
Planning Process: Modeling
• Mapped watts per square foot via CFD several times,
tweaking each time to determine placement of precision
air conditioners
• Room designed for 800-1,000 kW; CFD proves capacity
• 24-inch raised floor, multiple precision air conditioning
units
• White space to grow into: 40 to 50 percent allows for
doubling of capacity
• Rerun CFD every three to six months as changes are
required using actual values collected from thermistors,
branch circuit monitoring and other tools
Existing facility: Servers per Cabinet
Planning Process: Profile Planning
Cabinet Profiles
Best Practice: Technology Infrastructure
• 44U, 45” deep cabinets to allow for distribution of power
and low-voltage cabling
• Metered Power Strips
– Every cabinet has at least two strips, one for each bus
– Some cabinets with additional strip behind a static
transfer switch
– Each strip has colored tape to quickly determine
which bus
• KVM over IP in each cabinet
– Centralized
– Decentralized
• iLO switch in each cabinet ($38/Ethernet port)
Best Practice: Power System Design
New caterpillar
Generator
1000 KW
Dual Bus
Dual Power
BUS Remote BUS
Supply Servers
A Distribution B
Automatic Cabinet
Transfer
Switch 800 A
#1 Emergency
RSC Distribution
Panel
480v
70 kW A/C
Automatic 70 kW A/C
Transfer
Switch
#2
SPAWAR 70 kW A/C
Liebert 610 UPS Liebert Precision Dual Bus
225 KVA Power Center
70 kW A/C 480v Input 225 kVA
BUS Remote BUS Dual Power
Supply Servers
A Distribution B
480/ Output 480 Volt Input Cabinet
BUS B 120/208 Ouptut
Automatic
Transfer
Switch
#3
Building
Main
Electrical
distribution
panel
Best Practice: Dual-Bus EPO Systems
Best Practice: Fire Suppression
• Very Early Smoke Detection Apparatus (VESDA)
• Clean Agent Suppression system
– People-friendly
– Environmentally friendly
– Lower quantity of agent vs. FM-200
– Over 3,000 pounds of FE-25 agent
• Four zones
• Over 150+ detection points (above and below floor)
• Sprinklers were required, therefore pre-action with
double-interlock was installed
Best Practice: Remote Management
• Integrated Lights Out – allows remote control of servers
from anywhere in the world
• People-free facility benefits:
– Removed NOC Command Center from raised floor
– Minimizes human error disruptions
– Allows managers to monitor and control environment
remotely
Best Practice: Proactive Monitoring
• Followed some of ASHRAE’s TC9.9 recommendations for
monitoring temperature in racks/cabinets
• Temperature and Relative Humidity
• Branch Circuit Monitoring
• SiteScan I/O-32 and additional units
• Leak Detection
• VESDA
• Fire Suppression
• Fire Detection
• Liebert Environmental Control Systems ICOMs Control
Network
Best Practice: Labeling
• Mechanical equipment
• Electrical equipment
• Electrical receptacles
• Grid Coordinate
• Racks and Cabinets
• Structured Cabling
Tools: Rackwise/DCM
• Used tool to document existing inventories
• Tracking relocation from one facility to another
• Manage asset location by rack unit
• Provisioning limitations per cabinet
• Reporting capabilities
– Space and Assets
– Power and Cooling
– Cost and Capacity Planning
– Rack Assembly with Cabling
– Service & Warranty Contracts
– Department & Customer Information
• Long term objective: to integrate a change/configuration
management tool to track and manage assets
Tools: Rackwise/DCM
Tools: Computational Fluid Dynamics
• Allowed modeling the complete room for the initial 6
CRACs and full deployment of 12
Tools: Computational Fluid Dynamics
• Initial assessment indicates potential problems related to
placement of precision cooling units…
1 2
4 5
3 6
CRACs
Tools: Computational Fluid Dynamics
• CFD Program allows for virtual change - validating
proposed solution prior to installation.