Sunday, September 2, 2007

Using VERITAS Cluster Server To Reduce Data Center Power Consumption and Carbon Footprint by Data Center | Andrew Harrison

Why You Should Read This Paper

Effective data center power and cooling management is an area of increased focus as corporations struggle to reduce their power consumption and carbon footprint. So much so, that at Gartner's 25th Annual Data Center Conference in November 2006, data center power consumption was a headline issue.

There many reasons for this heightened interest:

1. Gartner research presented at the conference predicted that half of the world's data centers will run out of power and cooling capacity by 2008 1 . A 2006 survey conducted for power supply vendor Liebert found that 96% of existing data centers will run out of power by 2011. Gartner attributes this to increasing server and storage footprints and massive increases in power density per rack. Gartner also suggested that today, a single rack can consume more than 30 KW, while less than a decade ago this figure would have been between 2 KW and 3 KW.
2. Increasing energy prices have resulted in power bills taking a larger and larger percentage of the total annual cost of running a system. In some cases the 3 year energy costs for commodity servers now exceed the initial purchase price. In the midto longterm, energy prices are expected to rise faster than inflation. "Today, energy costs typically form less than 10 percent of an overall IT budget," noted Rakesh Kumar, research vice president at Gartner. "But this could rise to more than 50 percent in the next few years. Most CIOs would struggle to justify that situation to company board members."
3. Concerns about the effects of CO2 emissions on global climate change often couple with mandatory local emissions cap and trade schemes where companies are forced to pay for emissions made above their annual limit. Many large corporations have also joined voluntary emissions trading schemes, despite not being required to do so by any local regulatory framework.
4. Availability of benchmark data showing data center energy efficiency and best practice targets.
5. Data center power consumption is becoming more visible. In 2005 data centers accounted for up to 1.7% of the total power consumed in the US. Research by Lawrence Berkeley National Laboratory in Berkeley, California suggests that data center power consumption will continue to grow at 14% per annum consuming an ever larger share of the global energy market.
6. Lack of local generating capacity, common in California, has generated increased negative publicity and pressure on large energy consumers from environmental and political groupings. The situation is so pressing that enterprises increasingly find they cannot obtain additional power at any cost.

Reasons why power budgets are inexorably increasing:

1. Poor utilization of existing assets and an inability to reclaim unused storage and servers has led to corporations deploying ever increasing numbers of servers and larger storage arrays.
2. IT organizations have focused on price/performance at the expense of energy efficiency. For example, power supplies which convert AC main inputs into DC power for commodity server use never run at 100% efficiency. Some operate at efficiencies as low as 6070%. Because energy costs are not visible to IT buyers, they have been unwilling to pay a small premium to for 90% efficiency despite this improved efficiency paying for itself within the life of the server.
3. Increasing numbers of mission critical online applications require clustered servers and highly resilient storage. A cluster of two servers consumes twice the power as a single server and, except in a minority of cases, delivers exactly the same throughput.
4. Huge growth in the number of applications deployed by organizations and footprint of these applications has resulted in large numbers of additional servers and storage being deployed to support these new and bulkier applications.
5. Data centers, designed when energy costs were relatively low, traditionally deployed inefficient physical data center infrastructure components such as chillers, UPS's and power distribution grids wasting large amounts for energy before it arrives in the system racks.

Executive Summary

Symantec provides a range of data center infrastructure products that allow customers to optimize server based solution performance, increase utilization of servers and storage, and protect services and data from downtime and loss.

This Symantec white paper examines the impact VERITAS Cluster Server and VERITAS Application Director can have on reducing data center Power Draw, in turn reducing utility bills and the carbon footprint required to support the applications hosted in an individual data center.

Utility bill and carbon footprint reductions:

* The paper shows that deploying a simple 3 node N+1 cluster instead of 2 dual node active/passive clusters can reduce power costs by up to $438,000 3 over three years while reducing the carbon footprint by 1407 metric tons of CO2 and in turn reducing carbon offset requirements if offsetting is in place.
* Deploying a 9 node N+1 cluster instead of 8 dualnode active passive clusters can reduce power costs by up to $3,069,000 4 over 3 years while reducing carbon footprint by up to 9849 metric tons and in turn reducing carbon offset requirements by up to $98,000.

Additional benefits:

* Reduced Server hardware costs to support the same level of performance and availability for mission critical applications. Similar reductions in network and SAN infrastructure requirements.
* Reduced management costs by reducing the number of servers required to support mission critical applications.
* Increased server utilization, active/passive dual node clusters can only manage 50% utilization at best. Using N+1 clustering or Application Director increases utilization.
* Reduced physical footprint. Some data centers are running out of physical space. N+1 clustering typically frees up substantial space because of the size of the servers involved.
* The possibility of delivering High Availability (HA) as a service rather than multiple HA solutions each configured subtly differently and each managed separately.

Achieving all these benefits requires no additional Symantec product expenditures because they are a N+1 clustering is standard component of VERITAS Cluster Server.


Article Source : www.symantec.com

No comments: