||Add To My Personal Library
July 30, 2010
Vol.32 Issue 16|
Page(s) 14 in print issue
Smart Building Choices For Uptime
Designing Your Structures Intelligently Can Keep Your Systems Up & Running Reliably
Uptime is job No. 1 for data center administrators tasked with keeping business-critical applications humming along uninterrupted. The penalties for unexpected data center downtime can be quite detrimental to a business, from loss of profits to unexpected maintenance expenses to damaged business reputation.
• A key part of data center design is the location of power suppliers that can provide power reliably, effectively, and consistently. Discuss power requirements with power suppliers in the area to ensure there is available power to meet business needs.
• Airflow and cooling are two infrastructure elements that can be designed for success during the data center design and construction phases.
• Successful data center design involves a variety of disciplines, including IT and infrastructure systems design. Bringing these disciplines together early in the data center design phase enhances the chances for success.
The task of developing a data center with consistent uptime begins well before construction starts. Enterprises looking to build new data centers or remodel existing ones must keep in mind there are numerous decisions early in the building process that can improve ultimate data center uptime and availability. Making good decisions during the data center design and building phases will pay big dividends later on when the data center is fully operational.
The power entering the data center is, obviously, critically important to steady uptime. An unreliable source of power can cripple data center operations by introducing an unexpected variable that’s completely out of administrator control. Administrators engaged in data center design and construction must carefully analyze power considerations.
A primary power-related consideration is the need to ensure that the local utility is able to provide plenty of power reliably and consistently. Darren Bonawitz, co-owner of 1102 GRAND (www.1102grand.com), a data center in Kansas City, Mo., says administrators should talk with the electric utility company to ensure that the location they are looking at has adequate access to power not only for today’s needs but also to support future growth. A planned data center expansion can quickly get derailed if a local utility cannot supply the additional power required for expansion.
A critical part of the power puzzle is the design of the power infrastructure required to run the data center. Jim Neumann, vice president of marketing at EDSA (www.edsa.com), a developer of power analytics solutions, says a perfectly run data center is the result of meticulous upfront design and relentless fine-tuning once operations begin. Neumann says power analytics software that designs and optimizes models of complex electrical power infrastructures can help administrators mitigate potential problems early during the design phase of the data center.
When the facility opens, power analytics software can continually compare “as is” and “as designed” performance to predict and prevent electrical power problems. The key, Neumann says, is to plan an efficient power structure and then continually monitor the system.
Airflow & Cooling
Data centers are filled with servers, networking equipment, and other devices that generate heat as they operate. Thus, airflow and cooling are two crucial infrastructure elements that ensure equipment does not fail due to overheating issues.
Steven Leidig, manager of enclosure engineering at Emcor Enclosures (www.emcorenclosures.com), says one of the most important initial considerations to help achieve long-term uptime and availability is proper airflow. Nonstandard configurations, Leidig says, should consider thermodynamic loading and analysis early in the design. There are tools, both analytical and empirical, that can help with this upfront design.
Also, when considering airflow, engineers must consider all system impedances, not just the fan/blower base cfm rate when performing airflow calculations. For example, Leidig says, filters restrict airflow, so when sizing the orifices, take into consideration installed configurations that may include cable bundles or other obstructions sharing the airway.
Poor cooling is another potential problem that can impact data center uptime. Eric Karmazin, managed services operations manager for Jelecos (www.jelecos.com), says inadequate cooling can bring a data center to its knees and result in clumsy fixes, such as the placement of noisy box fans to circulate out hot air.
Karmazin says that during the design phases, administrators should carefully assess rack capacities to ensure the proper amounts of amps/watts per rack are consumed to adequately keep the equipment cool and heat dispersed.
Other Design Considerations
For a successful data center construction or redesign project, it is important to bring together all parties participating in the design process. For example, in order to help improve the ultimate uptime and availability of a data center during the building process, it is imperative to bring the facilities design company into the process very early, together with the facilities department and IT management, says Ian Jagger, worldwide marketing manager for Data Center Services at HP.
Then, there are a series of trade-offs that establish what the design criteria should be, including reliability vs. cost, high density vs. flexibility, and monolithic design vs. modularity, Jagger says. The higher the reliability requirement is, the higher it pulls energy, building, and operating costs. The key is to build along modular lines, as not all systems and applications are as critical as the next, he adds. Building pods and IT zones with different levels of appropriate availability lowers capital and operating costs significantly and delivers the right level of reliability while providing flexibility for capacity growth.
Other design considerations are to identify the criticality of the facility and applications hosted, design the uptime availability to meet the business risk, and carry out a fault tree analysis of the design, says Maurice Julian, UK facilities project director at HP. Design personnel should also design the infrastructure for ease of maintenance because overly complex systems lead to maintenance failures via human error.
by Sixto Ortiz Jr.
Is Colocation An Option? |
Rather than building brand-new data centers, many companies are turning to colocation arrangements. This alternative to building a new data center can present companies with numerous advantages.
For example, outsourced service providers can offer hosting with multiple carriers, so administrators are no longer tied to a single supplier and physical infrastructure, says Kevin Dean, CMO at Interxion (www.interxion.com). If one carrier’s system goes down, Dean adds, administrators can automatically switch to another, promoting resilience and ensuring availability.
Another advantage to colocation is the option to convert CAPEX to OPEX by outsourcing where possible and leveraging a shared infrastructure. Also, a company can spread the cost of using an enterprise-class data center while enjoying the flexibility of scaling requirements in line with business growth, Dean says. In this model, as a company pays for and uses what it needs, it also benefits from the inherent efficiencies in energy consumption that sharing an infrastructure with a number of companies brings with it.