2008 Strategies for Green IT at Penn:Data centers and servers:Final Recommendations

From Provider Wiki

Jump to: navigation, search

The information below are the final recommendations from the Green IT Data Center and Server subteam.

Contents

Determining Initial System Requirements

  • Can you compare your needed requirements to machines you already own?
  • Do you have spare compute/disk on your existing hardware to accommodate these needs or is new hardware necessary?
  • Can you get by with lower-voltage CPUs (they tend to have slightly lower performance)?
    • Is additional hardware even necessary?
    • Can you install a virtual machine on your own hosted server, or on a virtual hosting service (such as the one provided by Technology Support Services)? Costs may be cheaper over the long run going with a hosted service than by purchasing and configuring your own hardware.

Vendor Offerings

Based on the results of 'System Requirements', spec out similar machines from the major vendors. E.g. a 1U,2S,4core,16GB,2spindle machine with an FC card, something like a Dell PE1950, or an IBM x3550 or an HP DL360 G5 or a Sun X2100 M2.

  • Get quotes from at least three vendors for the same specs.

more green than usual

When spec'ing out your server requirements, consider the following options

  • Get quotes for low voltage CPUs and 2.5" disks.
  • Check if there's a markup and if the power savings will be worth it.
  • Especially when purchasing multiple machines, can you specify details like no printed manuals and no power cords?
  • Make sure that you are getting power supplies that meet Energy Star 4 requirements, which includes specifications for 80+ power supplies (power supplies that are 80% or greater efficient).

For more in-depth information on Energy Star, please see the Energy Star Information page.

Dell

Product Offerings

This product grid is as of 5/2008

LineModelForm FactorProcessorSpeedMax RAM
PowerEdge19501UUp to two Dual-Core Intel Low Volt Xeon 5148 2.33GHz16 GB@667MHz
PowerEdge19501UUp to two Quad-Core Intel Low Volt Xeon L53101.6GHz16 GB@667MHz
PowerEdge29502UUp to two Dual-Core Intel Low Volt Xeon 5148 2.33GHz16 GB@667MHz
PowerEdge29502UUp to two Quad-Core Intel Low Volt Xeon L53101.6GHz16 GB@667MHz
PowerEdge29702UUp to two dual-core Low Power AMD Opteron™ 2212 HE 2.0GHz 16 GB@667MHz

IBM

  • Blades: IBM BladeCenter HS21
    • no IBM gear on current Energy Star list?
    • IBM is encouraging virtualization to consolidate servers, thus reducing the load on data center supporting equipment.
  • IBM downloadable power configurator

SUN

Placeholder for information on SUN as a hardware vendor and their sustainability efforts.

Choosing a Physical Location

Picking a data center where Green IT practices are in place may be better than trying to host a small number of servers locally. In general, it is better to install your server in a larger data center than to try and set up a small space for a few servers. Infrastructure for providing power conditioning, cooling, switching, etc uses up a significant amount of power. If these resources can be spread over the needs of many servers (such as in a larger data center), the per-server costs are reduced significantly. While having physical access to your machines may be convienent, remote access is typically sufficient for most management tasks.

Considerations include:

  • Determining the power requirements of all related hardware
    • Firewall
    • Switches
    • UPS
  • Determining the power load of associated cooling equipment
    • Centralized data centers will have much more efficient cooling than smaller scale centers (or closets).


There's no one-size-fits-all solution for data centers. Every situation will vary depending on infrastructure requirements which include but may not be limited to:

  • Rack space for initial requirements and the need for future growth.
  • Power
  • Air conditioning
  • Networking equipment
  • Accessibility



Pros and Cons

Pros

Cons

Do It Yourself Data Center

  • You control your own data center environment.
  • Convenient access.
  • No need to run across campus to the data center
  • Up-front costs of building your own data center. This includes everything required to get the room ready for use which may include:
    • The server rack(s)
    • Networking infrastructure (Routers & VLAN, Hubs/Switches, Hardware firewalls)
    • Air conditioning costs (rack units, computer room units)
    • Possible power upgrades (more outlets, 120V/250V, additional fuse box)
  • Physical space for future growth might be limited.

Shared Hosting

  • Eliminate up-front costs associated with building your own data center.
  • Deploy within an existing data center infrastructure.
  • Shared power, cooling, network infrastructure is much more efficient
  • The overall space and load requirements are substantial.
  • Your data center requires intensive management.
  • Physical security is a major concern.
  • Possible data center fees might cost more in the long run.

Improving Air Management

An example of poor cabling
An example of poor cabling
An example of excellent cabling; Image From Purdue's HPCC Barn-raising
An example of excellent cabling; Image From Purdue's HPCC Barn-raising

Data Center Considerations

Improving "Air management" - or optimizing the delivery of cool air and the collection of waste heat - can involve many design and operational practices. Air cooling improvements can often be made by addressing:

  • short-circuiting of heated air over the top or around server racks
  • cooled air short-circuiting back to air conditioning units through openings in raised floors such as cable openings or misplaced floor tiles with openings
  • misplacement of raised floor air discharge tiles
  • poor location of computer room air conditioning units
  • inadequate ceiling height or undersized hot air return plenum
  • air blockages such as often happens with piping or large quantities of cabling under raised floors
  • openings in racks allowing air bypass (“short-circuit”) from hot areas to cold areas
  • poor airflow through racks containing IT equipment due to restrictions in the rack structure
  • IT equipment with side or top-air-discharge adjacent to front-to-rear discharge configurations
  • Inappropriate under-floor pressurization - either too high or too low

More Solutions:

  • Some of the solutions to common air distribution problems include:
  • Use of "hot aisle and cold aisle" arrangements where racks of computers are stacked with the cold inlet sides facing each other and similarly the hot discharge sides facing each
  • Blanking unused spaces in and between equipment racks
  • Employ Liquid Cooling of Racks and Computers Liquid cooling is a far more efficient method of transferring concentrated heat loads than air, due to much higher volumetric specific heats and higher heat transfer coefficients.
  • The most common current approach is to use a chilled water coil integrated in some manner into the rack itself. Liquid cooling is adopted for reasons beyond efficiency; it can also serve higher power densities (W/sf). Energy efficiencies are typically realized because such systems allow the use of a medium temperature chilled water loop (50-60°F rather than ~45°F) and by reducing the size and power consumption of fans serving the data center. With some systems and climates, there is the possibility of cooling without the use of mechanical refrigeration (cooling water circulated from a cooling tower would be enough).

Server Specific Considerations

In general, consolidated racking of servers, and keeping the associated cables as neat as possible will help encourage airflow, and lower overall cooling costs. Keep cables tied to the side rails and bundled together. See the images to the right for examples of good and bad cabling practices.

OS Optimizations

Operating System (OS) energy saving optimizations are based primarily on the software and hardware being used. Many OSs now utilize the dynamic frequency scaling capabilities of modern day processors by default. This built-in functionality allows administrators to focus on other issues in the data center.

The links below provide more information on these issues.

Monitoring/Metering

Keep an eye on your overall server processor load as well as your power consumption. Monitoring can be done with a switched and metered PDU (power distribution unit), with power-aware server management interfaces, or with a temporary clamp-on meter. A single, high-powered server running near capacity is more efficient than several smaller servers that are under performing.

End of Life

One thing to consider is whether or not you can extend the life of your existing hardware or if purchasing new equipment will get you more efficient by default. In general, newer hardware will have more efficient power supplies, processors, etc than older hardware.

Extending_Server_Life

Another consideration is proper disposal of your machines. Recycling options have their own pitfalls to watch out for. Penn has some options for consideration:

  • Be aware that there are specific EPA regulations with regards to the disposal of electronic waste
  • Penn's Center for Community Parnerships
    • They will refurbish machines that meet their requirements, and distribute them to organizations in the Philadelphia area.
  • Elemental Inc. are a local electronics salvage organization that will properly recycle and/or dispose of older equipment.

There are several other options that are listed on Penn's Facilities recycling website for recycling, and on Penn's Computer Connection page.

Ebay started the The Rethink Initiative as a way to raise visibility of various electronics recycling efforts by manufacturers, governmental agencies, and other participants. There is a comprehensive list on the site of existing efforts and options within local communities for recycling electronics.

Image:KermitFrog.jpg

Personal tools