Best Practices for Greening Your Data Center

Filed under: 1sdn,Green Computing | |

green data centerRising energy costs and a focus on Green IT mean that it is time to find ways of reducing energy consumption and cutting data center bills.  There are several best practices that will help guide you towards a greener and more efficient data center solution.

by Konstantin Gorshkov, Green Economy Post


When it comes to IT application and storage solutions, there are many ways to save energy and reduce your costs.  Whether you have a private data center that fills an entire building, a floor, or just one room, rising energy costs mean that the right decisions will yield a signification ROI for your organization.  Even if your IT is managed remotely through a hosting provider with co-location or dedicated servers, there are ways that you can reduce your organization’s energy consumption with some attention and the right technologies.  Most recently cloud computing has provided the option of dynamically scalable and virtualized IT solutions that allow you to reduce on-site data center needs.  Regardless of your IT scale and budget, there are plenty of best practices for reducing costs and increasing efficiency.

Create a Baseline

A great place to start is by tracking your data center’s energy usage to create a baseline.  If the electrical consumption of a data center is lost in the total bill for the building, then start keeping track with the involvement of the IT department.  Once you know how many kilowatt-hours the data center uses and how much it is costing, you can start looking for ways to improve its efficiency.  A data center’s energy usage consists mostly of server and HVAC electrical consumption. There are several methods that you can use to benchmark and track efficiency, including Power Usage Effectiveness and Corporate Average Datacenter Efficiency, as explained in the post How to Benchmark Datacenter Energy Costs from InfoWorld.

Make a Case For IT

There are plenty of different energy efficiency calculators that can help put things in perspective and build a solid case for greening your data center. A business case that takes the long-term view will help to create a convincing ROI. The 42U site (named after a type of server rack) provides both a PUE/DCiE calculator as well as an energy savings calculator.  It may help to compare your operational efficiency with other data centers of the same scale.  The Green Grid website is a great resource that can help inspire and guide your green IT project.

Right-size Your Digital Fleet

Your servers are likely always on and therefore using a great deal of electrical power and generating heat. Yet, they might be underutilized — active and processing requests only a small percentage of the time.  Some servers might be running applications that are infrequently needed or simply no longer used. There are plenty of tools that allow a system administrator to track server utilization, and this information can be a great addition to your baseline measurement toolkit. A decent server utilization percentage will significantly increase your data center efficiency.  Fewer running servers will reduce electrical and HVAC costs and can free up existing server hardware.  Server utilization can be increased by consolidating server functions and tasks onto fewer machines, by taking advantage of server virtualization, and also by using cloud computing services where demand is known to fluctuate.

Use Efficient Power Supplies and Servers

A great deal of electricity is lost in the conversion process, so you can reduce electrical consumption by switching from 110V AC to 220V AC, or by switching to 48V DC power supplies. Using more efficient power supplies for your servers can go a long way toward reducing power consumption.   Newer servers also have higher-efficiency processors with better cooling systems.  The built-in power management features of servers are seldom utilized — ideally, data center energy usage should fluctuate in step with demand.

Partition and Rearrange the Data Center

HVAC costs can be reduced by properly segmenting a data center so that empty space and unused server racks are not climate-controlled. There are ways of improving the efficiency of an HVAC system with smart layouts such as the Hot Aisle/Cold Aisle server rack configuration. In this setup each two rows of server racks face each other and cold air is pushed into the server rack fronts (the cold aisle) and hot air is pulled out from the backs (the hot aisle).  Proper air conditioner placement can reduce cooling needs by making the most of cool air before it dissipates into the room.

Check the Temperature

Data centers tend to be overbuilt with a full load of running servers in mind, so it may be possible that some of the air conditioner units can simply be shut off. Newer data center equipment can be run safely at higher temperatures and keeping the temperature at 68-72*F rather than 62*F will result in significant cost savings. However, according to network security and product development consultant, Matthew D. Sarrel, CISSP, it is important to check the manufacturer’s specifications on existing equipment before raising the temperature and monitor performance and availability afterward.

Build a Smarter HVAC System

If your data center has only air-cooled air conditioner units that are all constantly running, a water-cooled air handler with fans may be a better choice.  An air economizer is a great way to create a truly green data center, since it cycles outside air through the data center, whereas a traditional data center cools and recirculates the same air.  Minimal humidity control combined with vapor seals can reduce humidity control costs. A smarter cooling system with variable speed fans will certainly increase your data center efficiency. (See Five Strategies for Cutting Data Center Energy Costs Through Enhanced Cooling Efficiency ).  If the data center is located in an appropriate climate, a plate heat exchanger for a water-cooled system can take advantage of the cold outside air.

© 2010, Konstantin Gorshkov. All rights reserved. Do not republish.

Shortlink:

Protected by Copyscape Duplicate Content Finder
Posted by Filed under 1sdn, Green Computing. You can follow any responses to this entry through the RSS 2.0. You can leave a response or trackback to this entry
Line Break

Author: Konstantin Gorshkov (2 Articles)

Konstantin Gorshkov has followed developments in green tech for several years. He believes that technology can go a long way towards creating sustainable and energy-efficient bright green cities. Konstantin's engineering background allows him to analyze the latest developments in green news with realism and objectivity. He combines his penchant for writing and his knowledge of green tech here at the Green Economy Post. Follow him on Twitter.

  • Pingback: Tweets that mention Best Practices for Greening Your Data Center | The Green Economy Post: Green Careers, Green Business, Sustainability -- Topsy.com

  • http://silverline.librato.com sshalini

    Nice post!

    With reference to ‘Right sizing your digital fleet” section above:

    >>
    A decent server utilization percentage will significantly increase your data center efficiency. Fewer running servers will reduce electrical and HVAC costs and can free up existing server hardware. Server utilization can be increased by consolidating server functions and tasks onto fewer machines
    >>

    Industry reports suggest that avg typical data center server utilization is way below 15% Ref: http://bit.ly/c8RWnE. There is interesting analysis done by James Hamilton (Ref: Mix Talk http://live.visitmix.com/MIX10/Sessions/EX01), where he points out that servers are the biggest cost in a data center. More in-depth power, energy and server facts at his blog http://perspectives.mvdirona.com/.

    You make a good point about consolidating server functions.
    Would you have more insights on the kind of server functions that could be consolidated, real examples may be?

    In the server function (workload) consolidation context, I would like to suggest Librato’s Silverline technology. It helps harvest server resources (CPU, mem, disk and network IO) which are unused by primary function (or application) and allows secondary workloads to run without impacting application performance, requires NO changes to the application itself.

    Full disclosure: I work on the Silverline team at Librato.

  • Konstantin Gorshkov

    Hi sshalini,

    Thanks for your comment and links. Low server utilization percentages under 15% are all too common and frequently avoidable with the right attention and technologies.

    Most server functions are great for consolidation, but server consolidation needs to be properly planned to account for the types of applications and processes, as well as their demand profiles and redundancy requirements.

    The best candidates for consolidation are applications and routine processes that have a predictable and stable system load footprint in terms of cpu, network, and memory utilization. Other good candidates include internal and non-critical applications, rarely used apps and utilities, and mysterious legacy systems. Real examples might include an archival or backup process, an email service, intranet apps, and storage.

    Customer-facing applications, mission-critical systems, and apps that have a fluctuating or unpredictable demand or system usage footprint might not be the best candidates, or should be considered carefully.

    When consolidating, choose the best servers as the host system and carefully consider the performance and redundancy implications of having more apps and systems running on fewer physical servers. Keep in mind that server resources can be allocated more granularly across the apps that are now sharing the same server, or to the virtual machines running on the server.

    Consolidation, done properly, should increase your server utilization and lower energy usage without affecting performance, introducing new points of failure, or increasing risk. In fact, it might be a good opportunity to improve your IT redundancy strategy and app performance.