COOLING ARCHITECTURE


myriad of interrelated, non- linear factors which conform to raise inlet temperatures by drawing air back between racks or lower efficiency by requiring excessive work from server fans. Traditional design methods fail to take into account the properties of cooling architecture in a way that produces quantifiable predictions. Consequently, to ensure the desired cooling conditions are met, conservative estimates of CRAC volume flow and temperature requirements drive design, hindering precision. As part of the cooling process, the CRAC unit fans impart momentum upon the air, raising the static pressure, thus driving the air flow. From the fan affinity laws, where power consumed rises with the cube of the speed of fan rotation, it can be seen that if a fan has a 15 percent higher volumetric flow rate than necessary, 39 percent of the power consumption is wasted. It is clear that by increasing the precision to which cooling architecture is designed, efficiency can be gained. This precision, required for the most efficient designs, may only be achieved through a different design approach, developed entirely from scientific, physical laws with a minimum of assumptions.


THE BOTTOM-UP APPROACH


At a fundamental level, as a by-product of performing computational tasks, the individual components within a data centre continuously generate heat. Without any methods of transporting this heat, the heat build-up would cause computing equipment temperatures to rise continuously, increasing the risk of failure, until the equipment overheats and fails catastrophically. By transferring the heat to a material, and transporting this material away from heat sensitive equipment, a safe, steady temperature can be achieved, controlled by the properties of the material and its interaction with the computing equipment. Air is the obvious and most utilised material for this purpose,


ArchitectNews.co.uk | Architects Choice | 39


owing to its availability, heat absorption properties and the ease with which it can be moved by fans and ducts. Other cooling architectures utilise direct to CPU water cooling, as water has a higher capacity to absorb heat than air and is also easily directed with pumps and tubing. It is this ability of fluid to convect heat that underpins data centre cooling. For an air- cooled data centre, it is therefore advantageous to possess a detailed knowledge of the air within all parts of the data centre during all aspects of operation, as this describes the location of heat and thus the locations and causes of any inefficiency.


The scientific study of the movement of fluids is termed fluid dynamics, and fluid flows are described by the Navier Stokes equations, which are a set of non-linear, coupled partial differential equations. The mathematical complexity of these equations is such that no true, analytical solutions currently exist for many real


flows. Their complete solution requires an understanding of one of the major unsolved problems in physics, and will earn a $1,000,000 prize from the Clay Mathematics Institute. It is, however, possible to solve the Navier Stokes equations for complex, real flows, in a way that produces a solution precise enough for engineering applications, including: weather forecasting and the design of building HVAC systems, aircraft jet engines, hypersonic re-entry vehicles, and data centres.


This method of solution transforms the complex mathematics into an


approximate form that can be solved by computer processor, allowing the behaviour of the fluid to be predicted. The problem of fluid flow is now one of numerical analysis, a branch of mathematics concerned with finding approximations to mathematical functions, and their associated error, instead of exact solutions. This is the approach of Computational


Fluid Dynamics (CFD), a tool utilised within all industries concerned with the movement of fluids.


CFD APPLICATIONS As CFD deals directly with the underlying physics of cooling, it has wide applications for data centres. For design, CFD can be used to validate an existing design to prove the cooling architecture is fit for purpose, giving the rack inlet temperatures and CRAC running conditions, as well as illustrating any inefficiency and why it arises. Failure scenarios can be run in order to test the cooling ability of fewer CRAC units, or the interaction and mixing of standby CRAC unit flows with cooling flows from other sources, for use with Free Cooling.


CFD can also be used earlier in the design phase to drive the design. If the cooling architecture’s key


requirements are outlined, multiple designs can be tested and any variable (such as CRAC type and running point,


Figure 1 – sizing of CRAC units from IT load.


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52