Energy crisis: What does it mean for your data center?

Blade servers deliver unprecedented processing ability, but this increased performance comes at a hefty cost, literally - blade servers consume, typically, 10 times more total power than conventional rack servers, and require at least a four-fold increase in cooling capacity. This translates up to 3 to 4 times the total cost of ownership for the life of the asset.

Managing power costs remains a daunting challenge for IT administrators in all walks of life. Forecasting and budgeting server and storage power needs is tricky at best, especially in light of higher fluctuations in energy costs moving into 2007. Harnessing and conducting sufficient electrical and cooling power can surpass annual budgetary expectations just months into a new fiscal year, causing a cascade of cost adjustments and budget cuts to minimize the impact on operating expenses.

Unfortunately, the steady increase in utility power costs is well known to the oil and gas industry. This nationwide concern varies according to geographical location. However, in Houston, the cost of power for local data centers increased from 4 cents per kilowatt per hour to 11 cents - a 175% increase over the last 3 years. For companies that are running out of capacity in their existing data center, or those moving to blade servers, this unpredictability of power should be top of mind.

Building a data center during an energy crisis

The rising cost of power makes it nearly impossible to predict the expenses of retrofitting or green-fielding and operating a data center. Additionally, companies overlook and underestimate critical capital expenditures and operating expenses that can fluctuate over time. Further, there are misconceptions associated with the costs of both capital and operating expenses.

Capital expenditures

The typical capital expenditures for data centers include cost of design, engineering, switch gear, PDUs, HVAC, and electrical systems including monitoring, generators, and redundancy of all equipment. However, as rule of thumb, it costs 100% of the data center asset for installation and design.

For example, if a company purchases a generator for $300,000, it will cost $300,000 to design and install it, not including maintenance expenses. Further, if a company requires redundancy to ensure the availability of all systems, the cost of installing generators doubles.

Redundancy

All redundancy is not created equal. Most data centers have several single points of failure. Only in the past few years have high-availability data center infrastructures been designed and implemented with true two-end configuration.

The power diagram (Figure 1) illustrates a true, two-end configuration power infrastructure. The drawback of this can cost $5,000 per kilowatt for design and installation, not including maintenance.

null

Operational expenses

The given operational expenses include rent, maintenance, security, and electricity (kw/hour). What’s often overlooked is the cost of capital and depreciation. In planning 5 years out, a significant portion of the capital infrastructure will be un-utilized throughout the majority of the 5-year life of the asset.

These costs accrue on a monthly basis and appear at the bottom line of a company’s overall balance sheet. While high-level executives are exposed to this number, IT has no visibility into the actual cost of operating the company’s data center.

True cost of utility power

As a general rule of thumb, it takes $1 to cool every dollar in direct consumption in a data center. For example, a 250kw consumption rate would equal to $43,200 a month in utility cost or approximately $2,600,000 over a 5-year period.

Rising operational expenses remain a strategic wedge for IT leaders seeking to gain greater control over data center costs. These concerns point to outsourced hosting as the best means to achieve optimal return on investment, without sacrificing performance and flexibility.

Colocation provides a scaleable solution

Some outsourced data centers are built to accommodate extreme conditions and deliver reliable backup for even the most demanding server and storage environments.

Rather than absorb unpredictable power expenses in addition to the costs of deploying and managing an in-house data center, most enterprises find infrastructure hosting to be the best approach that provides the most predictability of cost and performance - with 100% uptime, while the host is equipped to respond to virtually any power outage or natural disaster.

Hosted data centers deliver cost stability and peace of mind. Outfitted with best-in-class power supply equipment, hosted data centers are ideally positioned to economize power whenever and wherever possible through optimal distribution of energy controls and resources.

This translates into predictable, budget-conscious hosting costs for the customer. Should fluctuations in energy costs persevere or worsen, the hosted data center absorbs the added expense.

Managing power consumption

A hosted data center can deploy dedicated blade server racks that deliver faster, more advanced processing capabilities using a far smaller footprint or square footage than conventional server racks. For the customer, this translates into greater performance per square foot. A high-performance center will position the racks on elevated flooring, meaning easier, safer deployment and modification of necessary wiring and cables, including secondary or backup conductors.

As high-density equipment becomes more prevalent, hosted data centers are mandated to find ways to achieve new efficiencies in electricity consumption. About 25% to 50% of data center costs are attributed to power consumption, and this number continues to increase dramatically and could surpass hardware and equipment costs in the near future.

This means that achieving optimal power distribution is paramount to hosted data centers. These centers are best equipped to achieve power efficiencies - most in-house data centers are not as capably outfitted, and would be less successful at outlining cost-saving equipment/resource configurations.

An optimal or baseline configuration for new server configurations is approximately 150 watts per square foot, with reserved high density zones up to 500 watts per square foot for dense computing hardware. This represents a substantial increase over non-high density zones, but again, high-performance hosted data centers are equipped to design and supply power configurations to accommodate these demands.

Adhering to compact power consumption guidelines ensures maximum efficiency per square foot within the data center, meaning customers use less real estate to achieve their results.

To put this trend into perspective, it costs approximately $200 per square foot to equip and deploy an N+ 1 data center running at 40 watts per square foot.

Taking maximum high-density consumption into consideration, it could cost a staggering $1,500 per square foot to build a data center running at 250 watts per square foot. For a 10,000-square-foot data center, this translates into an increase of $15 million in power and cooling infrastructure.

This is why many companies with in-house conventional data centers have opted to go with data hosting for its blade server and high-density hardware needs.

Dense computing experts

High-performance data centers are further equipped with best-in-class cooling equipment to keep high-density enclosures at an optimal temperature. Many IT managers overlook the impact of cooling output to managing a data center.

Sufficient cooling is absolutely essential in a data center environment, and especially with dense computing hardware, where cooling needs can be exponentially greater than those of conventional server racks. Should the cooling output be even slightly below the required levels, it can quickly result in overheating and shutdown as blade servers begin to absorb the heated air surrounding them.

Depending on configuration of the server floor, hosted data centers can apply cooling protocols in various ways. In an environment where high-density equipment is minimal, a borrowed cooling approach can be used, where temperature is regulated at an optimal level across the floor, with a rules-based protocol for diverting under-utilized cooling components to high density equipment.

In an environment where high-density equipment is more prevalent, a zoning approach is used, wherein high density servers are segregated from conventional servers and cooled separately, using dedicated equipment and controls.

Some customers use only high-density equipment and are zoned in a segregated layout, with only customer-specific dedicated equipment and resources. This creates a virtual “customer zone” within the hosted data center, a feature that some customers prefer.

Planning for the future

Whether looking to fully host its server environment or expand by hosting certain high-density components, a customer is advised to consider the following essential questions when developing its IT strategy:

High-density servers require up to 500 watts of power per square foot. You might ask yourself: Am I equipped to deploy and manage this workload? Are we able to weather unpredictable utility costs to power the equipment?

High-density equipment demands up to tenfold greater power - or more - to operate compared to conventional servers. This translates into exponentially higher infrastructure costs.

Before deciding on whether to build or increase in-house capacity or turn to a high-performance hosted data center to accommodate your needs, here are some more questions you need to ask:

• Is your current in-house data center maxed-out in terms of real estate?
• Do I retrofit the center to accommodate compact, high-density equipment?
• Is it feasible to expand the data center by adding more real estate?
• Do we have the budget to accommodate such a dramatic increase in power and infrastructure costs?

Once you know the answers to these questions, you can determine if your company is a potential candidate for utilizing a high-performance hosted data center to accommodate your IT and security needs

About the author

Blake McLane is the vice president of strategic business development for Houston-based CyrusOne [www.cyrusone.com]. The company specializes in providing data center for the oil and gas industry.