Trends 2011: Efficiency Without Compromise

By Gail dutton

Heading into 2011, businesses are dealing with a level of uncertainty not seen in at least the past 60 years. The challenges are immense, and data centers are in the difficult position of needing to grow to support the business despite tight budgets. Increasing efficiency without compromising operations is possible, but it’s not necessarily easy; and it’s certainly not instant. Here are some trends to keep your eyes on.

Integrated power management and analytics will be of greater importance, particularly as the Smart Grid takes shape. Cooling and energy usage will continue to be major concerns as prices rise. With an eye to the budget and maximum flexibility, new types of computing facilities are being introduced, and multiple systems are converging onto fewer networks. As a result, virtualization hacking has become a greater threat simply because virtualized systems are more common and harder for managers to track. Private clouds are being introduced to mitigate the security issues clouds pose.

Hang onto your hats. 2011 promises to be dynamic.

“One year ago, the number one issue for data center managers was efficiency,” recalls Peter Panfil, vice president and general manager of Liebert AC Power, a division of Emerson Network Power, headquartered in Columbus, Ohio. Now, those concerns begin with infrastructure management, heat density, availability and, at number four, efficiency, according to an Emerson survey.  Consequently, this year, data center managers must push the envelope for reliability and availability to find the optimum balance between management efficiency and operations efficiency.

Integrated Power Management
In today’s computing environment, almost everything from code to hardware is modular. The problem is that many of those modules are managed as silos. That’s particularly evident in the way the power infrastructure is managed, with IT power and building power managed by different divisions. The result, oftentimes, is a load imbalance that triggers outages.

Today, the barriers between IT and building power management are dissolving simply because they are becoming a barrier to efficiency and—worse—a barrier to competition, according to Steve Hassell, president of Emerson Network Power’s Avocent business, in Huntsville, Ala. Speaking at the autumn Interop conference, Hassell pointed out that virtual servers are spun up within minutes in power environments that are designed to adjust in weeks or months, sometimes overloading circuits, blowing fuses and reducing availability.

Data center integrated management (DCIM) tools eliminate that issue by providing real-time visibility to both IT and physical power systems, and prepare the way to leverage the Smart Grid as it is introduced. David Cappuccio, managing vice president and chief of research for the infrastructure teams at Garner, in Stamford, Conn., writing in the report, “DCIM: Going Beyond IT Problems,” predicts, “By 2014, DCIM tools and processes will become mainstream in data centers, growing from one percent penetration in 2010 to 60 percent.”

The best DCIMs will provide real time coordination among thousands of devices, allowing smarter, real-time decision-making in a dynamic environment, Hassell says. For example, a DCIM system will communicate with power, cooling, IT equipment battery monitors, and data collection devices in their own languages and feed that information to modules for inventory, configuration, operations, change, monitoring, facilities and access. “The idea is to have one repository that can be used from different areas…so users know the effect of changes before they make them,” he summarizes.

Cooling
High density cooling systems are replacing traditional hot and cold aisle configurations, in a direct response to high density computing.  Today’s usual rack densities of 7kW are expected to grow to 11kW by 2013 and to 17kW by 2021.  These increases, especially when added to virtualization, are creating hot spots that are best addressed at the source.

At the recent Chill Off 2 evaluation, experts from Lawrence Berkeley National Laboratories evaluated 11 cooling approaches. All focused on removing heat at the server or rack level.

These experts concluded that Clustered Systems’ direct touch cooling technology was the most revolutionary. This prototype replaces the server fans with chill plates that are cooled by liquid refrigerant. Called “unique” by Lawrence Berkeley National Laboratories, the direct touch cooling system transfers server heat to the server housing, then through the thermal interface material lining the cover, to a cooling plate chilled with refrigerant circulating through the plate’s micro-channel tubing. It is expected to achieve an 80 percent reduction in cooling costs compared to room-based cooling systems. With no fans, it is virtually silent.

The other innovative cooling approaches in Chill Off 2 were versions of rack coolers and rear door passive coolers using either air-to-water or air-to-refrigerant heat exchangers, and a server enclosure cooled with chilled water. Chill Off 2 results showed that passive cooling systems are more energy efficient than those with fans, that refrigerant-based systems are more efficient than those using water, and that rear door systems were more efficient than in-rack and in-row systems. Chill Off 2 evaluators also noted that refrigerant-based systems could run efficiently with warm temperatures of about 70 degrees Fahrenheit.

Other organizations also are improving their cooling strategies. Google, for example, is using an air wand that pipes cold air to specific components within its data centers, and several research institutes are developing liquid-based cooling at the device and even chip level.  To address concerns about evaporative (adiabatic) cooling, United Metal Products has developed an external adiabatic cooling unit that also provides humidity control—one of the main concerns about this type of cooling.

Energy
Although some data centers are installing renewable energy projects onsite, as a hedge against high costs and limited supplies, renewable energy isn’t cost effective and generally doesn’t supply enough power to run large data centers. Ken Brill, founder of the Uptime Institute, says “Typically, it’s difficult to get more than one feeder line into a data center. “The EPA is becoming increasingly stringent on local sources of pollution,” he adds, so fuel cells and other on-site generation options are becoming restricted. Likewise, “We’re seeing companies that installed onsite power generation programs turn them off after five years because that had insufficient engineering expertise and maintenance capabilities onsite,” to operate those projects efficiently.

As energy becomes more precious, power usage effectiveness (PUE) ratings are dropping in an attempt to conserve energy. A PUE of 2.0 is considered standard, but new facilities achieve closer to 1.0. Google’s data centers, for example, average a PUE of 1.2.

Analytics
Until now “enterprises have operated like a starfish, with several arms sprawling autonomously, and central control being very basic. But the days of sliding on sand are over. In 2011 that starfish will need to start tap-dancing,” forecasts Andrew Hillier, CTO of CiRBA, in New York City. To do that, data centers are embracing centralized analytics to provide a view of the entire IT operation so managers may craft the sort of well-coordinated approach that is used successfully by other business units.

This evolution of analytics focuses on workload placement and resource allocation to optimize data centers in ways that were not possible previously. “Centralized intelligence will enable more efficient and near-real-time coordination between different silos and IT stakeholders …so workloads can be proactively right-placed and right-sized,” Hillier elaborates. “More sophisticated approaches will be adopted, driven by expanding virtual footprints and the emergence of internal and hybrid clouds, where automation relies on accurate, broad-based decisions.”

Calum Smeaton, CEO of Sumerian, in Edinborough, Scotland agrees. “There’s too much data and not enough insight. Answering high-level questions— assessing the relative value of various IT projects, for example—can be difficult.”

Sumerian couples off-the-shelf tools with insights from economists and other specialists to maximize effectiveness. This approach requires a significant investment in tools, people and processes, Smeaton emphasizes, but the results are worth it. For example, when executing a right-sizing project for a European investment bank, Sumerian linked data from multiple silos to simulate different loads, identifying a business transaction footprint, heat map and applications that were being stressed.  “Load balancing helped them gain 40 percent (efficiency)” Smeaton recalls.

Limited Community Clouds
New cloud strategies are emerging as organizations try to balance expenses with business needs and opportunities. “Limited community clouds are very new…they’re just starting,” Guy Naor, CTO at Morphlabs, says from his Manhattan Beach, Calif. office. Unlike public clouds, which are available to anyone, limited community clouds share virtual resources only with select entities. For example, a limited community cloud may be shared by organizations using the same data center host, by a group of companies that work closely together or by non-competing organizations in the same industry. “The limiting factor is trust between entities,” Naor stresses, in determining whether limited community clouds will be effective.

Eventually, Naor envisions a fully distributed cloud that operates similarly to the power grid, with organizations around the world collaborating to maximize resource utilization, shifting loads based upon time zones. That vision requires fast and secure connections and a globally harmonized regulatory environment.

Facilities
When facilities need to be scaled up, modular data centers make sense. Giants like Google use rows of containerized data centers as the main facility, supported by power and cooling units. Smaller operations see them as a quick way to add extra capacity. Some are outfitted with solar cells to operate off the grid. Some of the newest entrants are better sized to accommodate server racks, and others are hardened against the electromagnetic pulses that weather experts and military minds say are inevitable.

But shipping containers outfitted as data centers aren’t the only modular approach available. HP introduced modules recently that look like a traditional data center. Multiple data pods are centered around a core support facility, offering the physical security of a traditional data center and the flexibility and quick addition of modules.

Yahoo!, in its newest data center in Lockport, N.Y., created another option, designing a pre-fabricated building that was assembled onsite in less than two months, cutting costs and minimizing construction time. That design can be replicated as it expands globally, too (see Case Study on Yahoo!).

Virtualization Hacking
“There are some serious pointers to 2011 as the year for virtualization and cloud-computing-specific hacks to make their way to headlines,” according to Johnnie Konstantas, vice president of marketing at Altor Networks, in Redwood Shores, Calif. “To date, these threats have been largely the stuff of academic analyses and are not associated with any widely reported security breaches,” she says.

The rationale for her dire prediction is the number of high-value assets that are being virtualized and recent cyber attack statistics. To counter the threat, Konstantas recommends automated security solutions, which also help counter human errors that are likely in highly dynamic environments.

Systems Convergence
Systems convergence is just starting, but will grow rapidly, according to Rod Sampson, director of enterprise solutions for Belden, in Richmond, Ind.  “How to get multiple systems onto internet protocols will be one of the main challenges for 2011,” he predicts. Organizations are migrating infrastructure, network, data and operational services from their own isolated cabling into integrated systems that carry most of the necessary services.   It’s not unusual for organizations to have eight or more separate systems for building management, security, entertainment, wireless, fire alarms, LANs and other things, Sampson points out. Bringing these onto a LAN increases efficiency by allowing system administrators to manage a few, rather than multiple, systems and eliminates redundant products, like firewalls.   “It’s easiest to begin this convergence at the infrastructure level, and then move to the network level,” Sampson says. The challenge is that the easy things have been done. Now audio and large data file sizes are being added to internet networks, stressing the networks. The other challenge is determining who controls the infrastructure. “It’s critical to get the facility engineers, IT and the end users at the table to discuss how to manage the infrastructure,” he stresses.

Conclusion
The trends that will shape 2011 have already begun, and their pace is only accelerating. Regulatory and fiscal constraints at national and international levels are being felt in all sectors of business, speeding the pace of changes as organizations scramble to stay ahead, or at least abreast, of conditions that call for less energy usage and better energy management, and increased security as well as greater flexibility. The days of operational silos are numbered. Instead, data center managers must manage systemically, taking a big picture view not just of their operations, but of how they fit into the entire organization.

Gail Dutton, a contributing writer to Data Center Management, covers the intersection of business and technology from Washington State. She can be reached at gaildutton@gmail.com