TECH TITANS BUILDING ROOM Randy H' Kartz

1
TECH TITANS BUILDING ROOMRandy H. Kartz

• EEDC DAC UPC
• By Ismael-Xavier-Gemma

2
Introduction

• Cloud computing
• Servers, network equipment and storage systems
  are spread all over the world
• A small interface (web browser) use their
  services from anywhere

MORE BIGGER DATA CENTERS
Total Cost of Ownership

• Facility cost
• Power and cooling costs
• IT and non-IT personal and equipment
• Software

3
The Problem of Energy Efficiency
Source HP, Data center cooling
strategies technology brief
4
Reducing Operational Costs

• Consume less energy and incorporate greener
  materials and processes
• Power management hardware
• Water-based cooling systems
• Scalability and Flexibility
• Innovation in computer architecture and software
• Examples, Future data-center locations
• Microsoft Quincy, WA. 43.600 m2 - 48 MW
• Yahoo Sunnyvale, CA. 13.000 m2 - ?? MW
• Google Mountain view, CA. (21)13.000 m2 -
  gt10 MW
• Siberia, etc.

5
Building importance
Google data centers location

• Warehouses physical location and design becomes
  a new degree of freedom (e.g. proximity to power
  facilities, water cooled systems, etc.)

6
Datacenter (1990)?

• Plug new devices
• Update cool system
• 1-2 MWatts

7
Datacenter (early 2000)?

• More than 100.000 servers
• Commonly 20MWatts
• Consuming 50 electricity and producing 20CO2

• In six years, datacenters will spend more money
  in energy than on equipment.

8
First solutions

• Microsoft Resize the power supply to the average
  demand
• Lets increase 30-50 more servers
• Better monitoring
• Google
• High-efficiency power supplies
• CPU dynamic voltage/frequency
• Energy saving of 20
• Others
• Improve AC/DC conversion (5-20)?
• New cooling techniques (50)?

9
Cooling solutions

• Air based systems
• Traditional solution
• Inefficient cold-air distribution
• Racks with air-conditioning units
• Smart cooling systems (louvers)Reducing 25-40
  cooling costs.
• Water based systems
• Evaporation method
• Cooling water with air

10
Cooling solutions

• Computing, power, and cooling systems in a module
• 16kW/m2 vs. 0.5 kW/m2

11
Results achieved

• Energy-efficiency metric
• Power Usage Effectiveness (PUE) facility's total
  power consumption/power used by servers, storage
  systems and network gear.
• PUE 1means the data center is using most of the
  power for computing and little power is devoted
  to cooling or lost in electrical distribution.
• PUE doesn't gauge the efficiency of the computing
  equipment itself.
• A 2007 study by the U.S Environmental Protection
  Agency (EPA) reported that typical data centers
  had a PUE of 2.0 or more.

12
Google today's results

• Average PUE of 1.21 for its six large-scale
  Google-designed data centers.
• The company says its efficiency improvements
  amount to an annual savings of 500 kW-h, 300 kg
  CO2 and 3785 liters of water per server.
• How Google achieves these results is a secret.
• They have patented a water-based data center with
  containerized servers floating in a barge and
  with power generated from sea motion.

13
Microsoft results

• They reported a PUE of 1.22 for the container
  portion of its Chicago data center.
• Microsoft is pushing the container design in its
  next-generation data centers with preassembled
  modules containing not only servers but also
  cooling and electrical systems.

14
Future trends

• Google engineers have called to develop servers
  that consume energy in proportion to the amount
  of computing work they perform.
• Servers consume as much as 60 of their peak
  power when idle!
• New sources to obtain electricity (e.g. solar
  power).
• The use of virtualization to consolidate servers.
• Servers at data centers run on average at 15 or
  less of their maximum capacity.
• Need for tools to automate the management of the
  servers, control their power usage, share
  distributed data and handle hardware failures.