|
 |
|
General Information
|
Add To My Personal Library |
January 1, 2010
• Vol.32 Issue 1
Page(s) 1 in print issue
|
Data Center Strategies For 2010
Plan For The Days Ahead
|
| Key Points
• The recession delayed many programs and much infrastructure growth, but continuing a policy of “buy nothing” may start to seriously harm enterprises.
• New methods of assessing value in the enterprise are likely to be based on TCO. This is why data centers are likely to increase their reliance on services.
• Energy savings will continue to be a top priority in 2010 as enterprises adopt numerous methods for reducing consumption. | | The past year isn’t likely to go down as one of the easier times most enterprises have endured, but the news heading into 2010 seems much more upbeat. Although that bodes well for data center budgets, it also means that there may be some adjusting and catch-up to do coming off of recession-period hunkering down. No matter what, change seems imminent, and the needs of 2010 will be different from 2009. Although none of us has a magic crystal ball, some industry insiders were willing to share their own insights into the coming year.
 Emerging From The Bunker
Much debate still swirls around the true state of the economy, and it’s hard to judge whether more budget pain awaits while still stinging from the past two years. Still, there would almost seem to be a logical end within sight.
“There’s going to be some lightening,” says Mesabi Group (www.mesabigroup.com) principal David Hill. “You can only tighten the reigns so far. You don’t want the data center to starve to death. If it doesn’t have enough resources to do the job, then you’re upsetting the growth of your company. You can cut back during a recession to get through it, but you can’t do that forever.”
In the past 12 to 18 months, we’ve heard from countless vendors, analysts, and enterprises that purchases were being deferred if not cancelled outright. But if the front office is doing its job, then the back office and infrastructure eventually have no choice but to keep up and scale.
“That deferred investment has now crept up,” says Steve Fink, director of the infrastructure service line at Avanade (www.avanade.com). “We have deferred functionality increases across corporate IT departments. Any expenditure that has happened must demonstrate a hard savings on capital expenditure. As a result, we’ve developed this habit of thinking that if an investment isn’t going to have a 12-month return on investment, it’s probably going to get deferred or cancelled off. That habit isn’t going to die anytime soon.”
Hence, IT managers are going to find themselves in the unenviable position during 2010 of having to add new infrastructure and support that has been put off as long as possible, yet will be subject to an unprecedented need for justification. Tolerance for capital expenditures will continue to be relatively minimal, but this is likely to engender new thinking on how purchases and programs are evaluated. Avanade’s Fink believes that organizations will need to carefully orchestrate IT processes to show that they’re really delivering service at an optimal or at least required rate. As an example, he points to compute resources and how most servers underutilize their processing capacity. Fink believes that enterprises in 2010 will adopt more and better tools for analyzing compute usage and planning appropriately for the expansion and contraction of the resources needed.
Naturally, the easiest way to grab compute resources on an as-needed basis is through service providers, and Fink strongly feels that service-oriented infrastructure will continue to grow in 2010, in part because it will help reduce spending on IT operational staff. If the services prove to be cyclic and/or predictable, then the process of continually approving those resources can be automated, requiring less time to execute activities and further reducing total costs.
However, achieving a better level of control may not mean adopting “best of breed” management strategies in 2010. According to Darin Stahl, lead analyst for Info-Tech Research Group, data centers have by now invested in tens if not hundreds of “best of breed” solutions for managing infrastructure. “This strategy actually moves the data center organization farther away from the lowest cost-to-serve or TCO utility infrastructure, significantly erodes any previous investment assumptions made on large infrastructure tools, and places in jeopardy the data center being able to manage and view the infrastructure end to end through the implementation of spot tools that don’t integrate without significant and redundant investment of time and money.”
A Different Shade Of Green
Additionally, Info-Tech's Stahl believes that reducing energy costs will continue to be a data center focus in 2010. The difference between past and present will involve finer levels of metering. Instead of monitoring power consumption at the infrastructure level, data centers will be monitoring at the wall and even at the rack.
“While monitoring power usage at a facilities level provides some basic insight,” says Stahl, “there is not enough information to allow the data center manager to implement targeted changes in the environment to control load. More granular instrumentation provides a more holistic view of the relationship between facilities power consumption and IT infrastructure load. By examining the relationship between power usage and IT services—applications, batch, backup, restores, etc.—it’s possible to modify the run times of events to smooth out the spikes that threaten to push the facility over capacity.”
Enterprise Strategy Group (www.enterprisestrategygroup.com) senior analyst Steve O’Donnell believes that the energy trends we’ve seen develop during the oil spike and recession will only continue and expand in 2010. The key drivers here are still-high energy costs, ever-increasing compute density in the data center, and the 1:1 relationship between compute energy and cooling dollars.
“A decade ago, data centers had a power density of about 50 watts per square foot, or 500 watts per square meter,” says O’Donnell. “Nowadays, we’re looking at 10 kilowatts per square meter—20 times more.”
The cost of dealing with that heat is increasingly becoming data centers’ leading headache, so an increasing number of technologies are rising to meet the problem in the coming year. Flywheel UPSes are far greener and generate less heat than their conventional battery-based counterparts. Economizers—systems that use cool outside air to drop data center temps rather than expensive air conditioning—are rising in popularity. Concurrently, hardware manufacturers are raising their ambient temperature specs so that data center managers can raise the set points of their facilities.
Yet another trend in energy saving is the move to computer systems based on DC power. The power supplies in servers convert AC power into the various DC voltages required by different components, but this is an often inefficient process that loses anywhere from 10 to 50% of the energy as waste heat depending on the power supplies and load specifics. Further conversions at UPSes compound the problem.
“We’ve actually got three stages of conversion,” says O’Donnell, “and at every stage you lose energy, and we deploy capital equipment that doesn’t need to be deployed. This is why more forward-looking data centers are starting to deploy direct DC-powered equipment. They bring AC in from the street, rectify it, charge up batteries, and then deliver the DC from the batteries straight to the computer equipment, taking out the UPSes and the power supplies—taking out two stages of conversion—and saving somewhere between 6% and 12% of compute energy.”
Server vendors have claimed heat savings of 15 to 30% from adopting DC-powered systems.
Gone But Not Forgotten
The past decade, and the past couple of years in particular, have brought many new tools and technologies to data centers, some of which can revolutionize an organization’s effectiveness and profitability, but some of the biggest gains remain to be had from the simplest back-to-basics concepts.
“One data center issue that’s gained a lot of traction heading into 2010 is maximizing the value of existing hardware and software assets,” says Charles King, principal analyst at Pund-IT (www.pund-it.com). “Consolidating applications and workloads through virtualization is a popular method for achieving this, but businesses can also profit from more mundane solutions, such as data center asset discovery and management. The only thing more wasteful than a misused tool is one that is unused, lost, or forgotten.”
As managers look forward into 2010, they should also look back and assess what some of the most successful strategies of the past were. Perhaps blending something old with something new is the way to longest-lasting success. 
by William VanWinkle
SME Quick Stats
According to recent Info-Tech Research Group survey data, 11% of enterprises will build a new data center in 2010 and 33% of enterprises will renovate or refresh existing facilities. Interestingly, more than half of the facilities projects (56%) involve data centers or server rooms of 1,000 square feet or less within small enterprises. |
|
|
