Theorem of Operation

KyotoCooling changes many long held preconceptions about cooling for many people. KyotoCooling is based on sound physics and engineering and represents the lowest risk option for a data center designer or owner. KyotoCooling is about control, nearly absolute control of the data center cooling situation, irrespective of changing loads and environmental conditions, all without need for operator intervention. In addition this total solution is tested and validated as a complete solution, as a product, with person years of development, and repeatable, measurable and discernible results. Finally this system is designed for deployment in continuous availability situation with no single point of failure approaches. Let’s examine how and why this is so.

First of all we begin with something simple. To cool with air you must control that air. This is intuitive and sensible enough, but what do we mean by control? We mean that we do not want to deliver more air then needed to cool with maximum efficiency. We mean that we want to not have that cold air contaminated with hot exhaust air. This recirculation or temperature contamination is responsible for the inconsistent temperatures from floor to top of rack in today’s data centers. Normally bypass and recirculation account for a 30% overall loss of efficiency in a data center. So we begin with the idea that we want to separate cold and warm air optimally. This can be by aisle containment, or chimney cabinet or fixtures for high density cabinets provided by a myriad of companies specializing in this control process.

The impact of this control is immediate and sensible. There is consistency in temperature of the air delivered to the IT systems in the data center. There is certainty in air flow. There are no longer anomalies and inconsistencies in temperature at inlet in the data center. We have resolved the first risk of cooling, knowing that we have a temperature that is constant across long aisle spaces from floor to top of rack. Now we can prove to you both by empirical observation and by computational fluid dynamics and most importantly through laws of physics and engineering why this works. We do not expect you to believe without first asking hard questions. We have the answers.

Traditional cooling paradigms fight the air flow of servers. They overflow, or overspeed, or underflow. We believe that it’s best not to attempt to circumvent or change the design conditions for servers. The engineers of those systems start with an assumption that they will have no static pressure difference from front to back. Overflow reduces ?T. This reduces the efficiency of any and all cooling processes. Underflow requires servers to do the work of downflow units. That is starving servers results in higher internal temperatures and fans working harder to pull air that they should not have to pull. Overspeed refers to situations where velocity of air from CRACS is too high for a cabinet in path to receive air from. The effect is typically to pull air from unplanned sources (recirculation) resulting in higher not lower temperatures.

We precisely control the flow of air based on measured ?T, and air flow differentials using a monitoring system known as Airlull. We are able to maintain precise balance in pressure between cold and hot spaces created by containment systems to deliver to servers a perfect environmental condition. We do not underflow, nor overflow. We deliver a precise volume of air at a precise temperature to the cabinet face, even under changing load conditions, all without operator intervention.

The next question is how to cool this. Here we start with the traditional cooling process. We provide an adequate amount of direct expansion or chilled water cooling to cool the intended load. Both are well understood processes. Both are able to be tested and validated and understood from a physics and engineering perspective. We design this in traditional ways using well understood and proven physics and engineering. Air has a specific capacity based on elevation to transfer energy. The amount of air required to unload an amount of heat can be determined by the permuted equation Q=mCp?T. This means that for a given Technology load in the data center we can determine the fans, motors, compressors and coils required to achieve the desire result. These selections are based on certified performance curves of the manufacturers of these subcomponents. These can be checked and validated at a subcomponent level. In addition the sum of the parts can be tested and validated in our integrated working solution. We would offer that since we are in control of air and this conventional cooling engineering is being applied that we have eliminated the most basic risk that the system will cool the data center because it has sufficient capacity. This however only makes us as good as the best cooling systems already in existence.

The next thing that we do is to add into the air flow a novel and patent pending design use of a heat wheel. This is not a conventional heat wheel design. This heat wheel does not bring outside air into the data center. That would mean a risk. We have designed a system that uses the heat wheel for air to air heat transfer without bringing outside air into the data center. Once again the physics and engineering is understood and performed with repeatable and verifiable results. We produce subcomponent selections and then test the total system for dynamic loss. The total system has been verified by external sources and long term tests in our research facility have borne out the effects over a wide range of ambient and load conditions. This nets to knowledge and assurance that the system acts as anticipated.

These heat wheels operate at extremely low rotational speed, and have performed reliably in industry for 20 years or more under harsh operating conditions. They are epoxy coated marine grade aluminum and impervious to normal environmental conditions. There is little to go wrong here. The wheel is rotated by a one and half horse power motor. Mean time to repair is under thirty minutes. While the wheel is being serviced the KyotoCell maintains full cooling capacity under direct expansion or chilled water design.

The KyotoCells are deployed in a redundant N+1 or N+2 design. This assures continuous service under maintenance or service with maintenance and under failure condition. There is high availability design within each KyotoCell, with up to twelve independent compressor steps, six high capacity backward inlet fans, quad coils, independent controllers on all subcomponents, redundant sensors, vibration monitoring for predictive bearing failure, filter change and compressor leakage monitoring and much more. A KyotoCell is designed to operate continuously without outage for years at a stretch.

Having said these things we are only scratching the surface of the KyotoCell. The sum here is greater than the parts and that is embodied and made true in our controls environment. This is the most sophisticated data room cooling control package in existence today. With over 240 IO points per cell this system is the concert conductor of the interplay of the parts. The system is a control the controller design based on a rigorous fault matrix of possible scenarios. Each component has between it and the Kyotocell controller an independent controller that ensures the continued operation of the unit during Cell level controller failure or changeout. At no time can a loss of any controller or subcomponent remove the entire capacity of the cell from the data center.

This Java based controller package is based on an open development and integration environment. The control code is protected by escrow and is delivered in a readable form. The development environment is NiagaraAX. This open architecture environment includes a rich set of certified object based interface controls modules and kernel.

Each Kyotocell is independent of all other cells, just as each subcomponent is independent of the cell level Controller. No single point of failure exists. The entire code package is tested and proven across many person years of development and validation. No site level programming is required, simply option selection. The system is self tuning, autonomous across load and environmental changes and component adaptations. The system is designed to act and react as necessary to achieve stable operating conditions. Load changes are automatically stabilized to target set point. This means the addition of load can be accomplished within minutes, irrespective of data floor location, automatically and without risk or disruption.

When a failure in a subcomponent occurs the system attempts to compensate using a classic failure mode model. In an N+1 or N+2 deployment, Uptime Institute Tier 3 and Tier 4 designs are achieved. All contingencies are planned for and dealt with in advance. No operator thought or intervention is required. Of course active alerting with prioritization is incorporated. This is a complete cooling solution that includes all possible scenarios pre-programmed and pre-determined. This is being in total control of your data center cooling.