A recent study out of Stanford University says that 1 out of 3 data center servers is a zombie. netTerrain Environmental Monitoring Module helps you fix that. Find out how - with real-use scenarios and savings calculations. This is a slide deck from a recent Graphical Networks webinar.

1. Drive New Levels of Efficiency in the Data Center
Fight Zombie Servers

2. Agenda
Slides:
The Zombie Epidemic: Overview
The Problem You May Be In
A Solution That Will Help
Real Use-Case Scenarios with Cost $ Savings
Live Demo
Questions & Answers

3. Zombie Servers:
It’s an Epidemic
Zombie servers – if they aren’t already
lurking around your data center, chances
are they’re on the way. Take a look at
how widespread they are.

4. What You Need to Know
30%
Of servers are
zombies
$$ ● 1 in 3 Servers is
drawing power but not
performing any function,
according to a recent
Stanford study
● Utilization of servers in
business & enterprise
data centers rarely
exceeds 6% - according
to McKinsey & Company
10
Million
zombie
servers
worldwide
//////////////////////// The Problem:
1 in 3 Servers is a Zombie

5. ////////////////////////
Excess Server Capacity On average, servers
only deliver 6% of their
maximum computing
outputNot only are 30% of servers
“comatose”, but servers in
business and enterprise data
centers today are underutilized.
In fact, on average, servers
deliver only 6 % of their
maximum computing output,
according to research via
McKinsey and Company.

6. 0
10
20
30
40
50
60
70
80
90
100
Before netTerrain
EM After netTerrain
EM
Series 1
////////////////////////
The Zombie Resistance
netTerrain DCIM-EM &
netTerrain Enterprise-EM,
the overall data center
experience is a marked
decrease in power
consumption and cooling
resources up to 30%.

7. ////////////////////////
1 in 5
Data centers are relying
exclusively on thermal
sensors and
spreadsheets to
maximize cooling
efficiency
////////////////////////
Are you relying on
manual methods?
If so, you’re not alone. 43%
of data centers today,
according to a recent
Redshift survey, rely upon
manual methods for capacity
planning and forecasting.
Don’t be scared: be prepared!

8. Powered by Intel DCM Energy Director
netTerrain Power Monitoring
• Real-time power and thermal
data for racks/blades
• Policy-based power capping
for racks/blades
• Aggregated & historical
control
• IT device power (PDU, UPS,
Network, Storage)
• Cross-platform support

9. How Does It Work?
netTerrain Management Console
DCIM Middleware
Monitor. Control. Trend. Scale. Standardize.
Hardware Protocols
NODE MANAGER PMI. iDRAC IPMI. iLO/DCMI IPMI.
IMM/IMPMI.
CMS HTTPS/WS-MAN.
OA SSH/CLI.IMM SSH/CLI. SNMP.

10. save you time and money?
How will netTerrain
Environmental
Monitoring
• Replace expensive smart power strips
• Streamline capacity planning
• Identify zombie & underutilized servers
• Measure your energy usage by device
• Easily identify power/thermal failures
• Power-aware VM migrations
• Power-aware job scheduling
• Continuous operation during power outages
• Improve your thermal profile
• Optimize your applications’ power usage

11. Zombie Devices
Use Case #1
ROI Formula
Key Takeaway:
Save power by decommissioning devices that have been drawing power but not
serving any function.
Assumptions
Power/Money Saved
(number of servers) x (percentage of ghost servers) x
(power consumed by server per hour) x (hours in day) x (days in year) x (electricity cost)
1. 1500 servers (100 Racks)
2. Ideal server consumption per hour = 150W or 0.15KW
3. Percentage of ghost servers as candidates for consolidation: 10%
1,500 x 0.10 x 0.15 x 24 x 365 x 0.1 = $20K
Use Case #1

12. (smart PDU cost – regular PDU cost) x (PDUs per rack) x (number of racks)
1. 100 Racks
2. 2 Smart PDUs/each rack
3. Cost of smart PDU = $1,000
4. Cost of regular PDU = $600
Use Case #2
Key Takeaway:
With DCIM, you get device –level power data and thermal data = no need for Smart
PDUs..
Power/Money Saved
($1,000 – $600) = $400 x 2 x 100 = $80K
Avoid the Cost of Smart PDUs
Use Case #2
ROI Formula
Assumptions

13. Key Takeaway:
Cooling costs are lowered by raising the room temperature through continuously
monitoring devices for temperature issues.
(percentage of power bill) x (# of servers) x (power cost/server/year) x (assumed
average PUE)
1. Percentage of power bill impacted per degree: 2% (industry standard)
2. Power cost/server/year: $263 (0.3 x 24 x 365 x 0.1)
3. Assumed Average PUE:1.5
4. Degrees that can be easily raised: 1°C
Power/Money Saved
0.02 x 4,000 x 263 x 1.5 =$31,560
Reduce Cooling Costs
Use Case #3
ROI Formula
Assumptions

14. Key Takeaway:
Proactive monitoring of real power consumption (vs. nameplate power) helps you
oversubscribe racks and increase your rack and space density.
(Cost per RU) x (# of Rus saved) x (# Racks)
1. Average # RUs per rack is 40
2. Average cost per RU (Tier 3 Data Center): $1200
3. # of Racks: 100
4. # Rus consolidated: 1 (conservative) Power/Money Saved
2 x $1200 x 100 = $120K
Increase Rack Density
Use Case #4
ROI Formula
Assumptions

15. Live Demo
Up Next:

16. Thank You
Questions?
814 W Diamond Ave #102,
Gaithersburg, MD 20878
USA
(240) 912-6223
Fred.Koh@GraphicalNetworks.com