The threat of overheating servers and/or critical equipment represents a potential crisis that regularly keeps engineers, IT professionals and owners up at night. Despite widespread awareness of this problem, finding a long-term solution for the delivery of healthy and dependable data centers remains elusive.
The industry has made several preliminary attempts to unify its approach toward data center cooling, such as standardizing hot/cold aisle cabinet layouts, front-to-back cabinet air flow protocols and vertical underfloor (VUF) and vertical overhead (VOH) air distribution systems, yet much work must still be done.
Here are some strategies that can be employed to help minimize these risks and further improve the standards for delivering healthy data centers that will function efficiently over the long haul:
- Design with flexibility in mind: Since most mission-critical facilities will undergo four or five complete computer equipment change-outs over their lifetime, they require a flexible design that will help to ensure minimal cost and risks throughout the upgrade process, while still maintaining continuous uptime. Some of the key factors in design flexibility include installing oversized piping and providing space for additional computer, mechanical and electrical equipment.
- Interface in the design process: IT professionals must interface with design engineers when planning the layout for the computer room. This will enable them to better understand how to lay out equipment racks applying the hot-aisle/cold-aisle strategy, and how to arrange perforated tiles (where a raised floor supply plenum is utilized) in order to optimize the cooling air distribution. Additionally, this interface will help the IT staff understand the limitations of the space which will be left in their hands long after the design engineer is gone.
- Use published guidelines: Several professional groups, such as the American Society of Heating, Refrigerating and Air-conditioning Engineers (ASHRAE), 7x24 Exchange, and the Uptime Institute have begun publishing universal standards to aid in space planning, estimating load growth, and designing mechanical and electrical systems. These guidelines represent the first step toward providing a benchmark for establishing higher standards in data center design.
- Employ Computational Fluid Dynamics (CFD): CFD modeling works in the following way: The design engineer describes the computer room by inputting a description of the physical space, including walls, roof, raised floor, equipment placement, and equipment loads. The CFD software then divides the computer room volume into millions of "cells." Energy, mass and momentum conservation equations are then solved simultaneously by an iterative process for each cell, with the result being the conditions of state - temperature, pressure and velocity – for each cell. The physics and mathematics behind this process are well understood, and the process provides good predictions for actual applications.
While it is important to employ CFD modeling during the data center's initial design phase, this process should also be re-applied throughout the facility's entire life cycle, from its initial design through its various upgrades and expansions as a proactive means of thwarting the overheating of equipment and unwanted downtime.
- Explore alternate system designs and layouts: In order to mitigate hot spots in the cold aisles, data center designs must also explore alternative designs and layouts, such as overhead air distribution and built-up air handlers following the use of CFD analysis. A good method for comparing the effectiveness of alternate systems is to use several available metrics that quantify the effectiveness of the cooling system.
The Supply Heat Index and Return Heat Index (SHI and RHI, developed by HP) helps document the extent of recirculation of hot air into cold aisles and bypass of cold air into the return of the air conditioning units. The Rack Cooling Index (RCI, developed by Magnus Herrlin) documents conformance of the cooling system with any given environmental specifications.
While the threat of overheating continues to loom large among the IT community these five steps can play an integral role in moving toward more standardized and cost effective means of designing and maintaining healthy data centers at their inception and throughout their entire lifecycle.
About the authors: Christopher Johnston, Vice President, Syska Hennessy Group, leads research and development efforts in critical and hypercritical facilities. Vali Sorell P.E. has 20 years of experience in conceptual and detailed design, project management, performance testing and commissioning, and business development for Critical 7x24 Facilities.