Green IT for Data Centers: Efficiencies through modeling, hardware selection and virtualization
When the United States Geological Survey asked for ways to expand their storage needs, computing power, and overall service capacity to make their Earth Resources Observation and Science (EROS) Center a Tier II (Uptime Institute) facility in the frozen plains of South Dakota, most people thought the costs would be enormous. However, through the smart use of software modeling, hardware selection, and virtualization, our team was able to provide one of the largest computer complexes in the Department of the Interior with a series of short-term, long-term, low-cost, and capital improvement overhaul options.
The presentation will be a case study of how to use graphic static flow modeling to optimize cooling efficiencies to save energy and operating costs, as well as looking at energy conscious substitutions of hardware and how virtualization can help increase the bottom line.
This presentation is geared for anyone that owns or operates a data center and wants to reduce their costs, increase their efficiency, and provide solid evidence with some golly-gee-whiz pictures to impress their bosses to open the purse strings for new projects.
1. Green IT Data Center Case Study:
Modeling, Selection, and Virtualization
Or “What I learned on summer vacation in The Matrix...”
2. Presentation Summary
Status Quo
Reduction Pressures
Growth Pressures
Modeling
TileFlow
Low-Cost/No-Cost Changes
Phased Approach Placement
Hardware Selection
Energy Star
Right-sizing – Cost/TB
Virtualization
3:1 or 15:14
2
2
3. Red pill or Blue Pill?
Status Quo – A Place to Start, not end
Existing
157 racks – 485 kW (99% heat steady state)
Tier 1 Data Center
Constraints
UPS
Electrical
Mechanical
Space
PATH OF LEAST RESISTANCE
Extrapolation
Cost/sf ~ kW/sf
Expansion – Add 230 racks
Loss of mission?
Reliability?
What EROS got instead
Reduction
638kW avoided - $202,881 saved annually
$615k cost avoidance
Tier 2 Data center
Master plan
3
3
4. Am I Bill or Ted? Wait, nope, I am Neo…
TileFlow Modeling
Investigation
Understanding constraints, needs, and existing conditions
Modeling
Input status quo into system
Determine “Hot Spots”
Average Temperature range
Under Floor Velocities
What happens if redundant unit fails?
Shift racks
Increase plenum space
Increase/decrease floor perforations
Install Overhead Return air
Install CRAC unit redundancy software
Hypothesize additional rack space
Engineer highly likely variables and confirm (WAG)
4
4
5. Am I an Elf or An Agent?: Agent Elrond
Phased Approach
Lessen Mission Impact
Reduce First Costs
“Just-In-Time” or Extra Redundancy
Depending on Expectations and Budgets
Allows technology improvements
Phase 0 – Low-cost/No-cost (Buys Time)
Phase 1 – Replace small HVAC (Prepare for Upswing)
Phase 2 – Move CRACs (Let projects finish, then move)
Phase 3 – Upgrade large HVAC (Big savings)
Phase 4 – Install “nice to haves” (additional savings/reliability)
5
5
6. Loop it!
Hardware Selections and Virtualization
Energy Star Rated Hardware
75% percentile
80% more efficiency minimum
– Idle Power Consumption – 50% standard
– Operational Power Consumption – Drop-in replacements for older power
Solid State Drive (SSD)Technologies
Multi-Cell Levels
Enterprise Management Systems
Virtualization
3:1 ratio potential
15:14; due to accounting/project constraints
Right Sizing
Cost/PB
6
6
9. Difference between knowing the path and walking the path
Other avenues
Set house in order
Tighten up envelope
Infiltration
Reflectivity
Get Reliability Up
Then think about options
Solar
Hydro
Wind
Human Potential
9
9
10. Was Ist Lost? Soul Patrol…. Free your mind
Conclusion
Status Quo
Reduction Pressures
Growth Pressures
Modeling
TileFlow
Low-Cost/No-Cost Changes
Phased Approach Placement
Hardware Selection
Energy Star
Right-sizing – Cost/TB
Virtualization
3:1 or 15:14
Meet Goals, Expectations, change the world, or at least you little piece of it.
10
10
11. What we do now is
up to you…
jeff.riegle@amec.com
612-252-3667
Any More
Questions?
Hang around
after.
Thanks to Innovative
Research, StellCo
Industries, and the EROS
Data Center Staff
Shaping the Future