Unit 2 Chapter 4: Cooling PDF

Unit 2 Chapter 4: Cooling  Index Cooling Costs – Power Cost – Causes of Cost – Calculating Cooling Needs Reducing Cooling Costs – Economizers – On-Demand Cooling – HP’s Solution Optimizing Airflow – Hot Aisle/Cold Aisle –...

Unit 2 Chapter 4: Cooling  Index Cooling Costs – Power Cost – Causes of Cost – Calculating Cooling Needs Reducing Cooling Costs – Economizers – On-Demand Cooling – HP’s Solution Optimizing Airflow – Hot Aisle/Cold Aisle – Raised Floors – Cable Management – Vapour Seal – Prevent Recirculation of Equipment Exhaust – Supply Air Directly to Heat Sources – Fans – Humidity  Adding Cooling – Fluid Considerations – System Design Datacentre Design – Centralized Control – Design for Your Needs – Put Everything Together  Cooling With any amount of power comes heat, and if there’s too much heat in the datacenter, you can expect trouble.  Cooling Costs: – Some estimates state that cooling can account for upward of 63% of your IT department’s power usage. – Figure out how much actually spending and how much actually need to spend. – How much power costs and how those costs are computed. – Electricity is paid for per kilowatt-hour (kWh). – This is a measure of the hourly consumption of electrical power. – A 100-watt (W) bulb uses 100 watt-hours of electricity in 60 minutes. – As such, ten 100 W light bulbs will use a total of 1 kWh of electricity per hour. – But electrical power costs are different around the country – International Data Corp. estimated that companies worldwide spent about $29 billion to cool datacenters in 2007, up 400% from 2000—IDC, 2006.  Causes of Cost: – Cooling is a major component of power consumption and IT budget. – Typically this number is expressed in British Thermal Units (BTUs) or kW. – One kilowatt is the same as 3412 BTUs. – Issues driving up power consumption and cooling costs, includes the following: Increased power consumption as more servers and storage devices are deployed. Increased heat density in the racks because of increased computing power in a confined space.  Irregular heat load in the datacenter. This is exacerbated by poor planning for heat management as the topology of the datacenter changes. Increasing power costs. A tendency to overcool datacenters. The “flood-cooling impulse” leads datacenter managers to overcool their datacenters by more than two and a half times what is needed.  Calculating Your Cooling Needs: – All the equipment in your server room generates heat. – Also the lighting and the people working there. – All these sources of heat contribute to the heat load of the server room. – In order for your air conditioner to cool a room, its output must be greater than the heat load.  – To determine the heat load consider following: 1. Room Size: To calculate the cooling needs of the room, use this formula: Room Area BTU = Length (meters(m)) × Width (m) × 337  2. Windows: Most often, server rooms have no windows. If you have windows, look at these formulas to determine which is most applicable to your datacenter: South Window BTU = South Facing Window Length (m) × Width (m) × 870 North Window BTU = North Facing Window Length (m) × Width (m) × 165  If there are no blinds on the windows, multiply the results by 1.5. Windows BTU = South Window(s) BTU + North Window(s) BTU  3. People in the Room: Total Occupant BTU = Number of occupants × 400  4. Equipment: Find the equipment’s power consumption in its documentation or on the vendor websites. Equipment BTU = Total wattage for all equipment × 3.5  5. Lighting: Lighting BTU = Total wattage for all lighting × 4.25  6. Total Cooling Requirement: Total Heat Load = Room Area BTU + Windows BTU + Total Occupant BTU + Equipment BTU + Lighting BTU  Reducing Cooling Costs It is wiser to deploy equipment that won’t chow down a lot of power. Equipment to use that can save money and help supplement environment.  Economizers: – Winter provides an opportunity to enhance cooling system by using the cold outside air to cool things down. – Employ an economizer for this. – There are two types: 1. air-side economizers and 2. waterside economizers.  Air: – An air-side economizer regulates the use of outside air for cooling a room or a building. – It employs sensors, ducts, and dampers to regulate the amount of cool air brought in. – The sensors measure air temperature both inside and outside the building. – If it notices that the outside air is suitably cold enough to cool the datacenter, it will adjust its dampers to draw in the outside air, making it the main source of cooling. – This cuts or eliminates the need for the air conditioning system’s compressors, which provides a big cost savings. – Because the economizers are drawing air in from outside, pollution can potentially enter the datacenter. – A larger concern is the change of humidity in the datacenter. – What you spend on filtration and humidification might be more than if you just used your regular air-conditioning system.  Fluid: – A water-side economizer utilizes evaporative cooling (usually provided by cooling towers) to indirectly produce chilled water to cool a datacenter when outdoor conditions are cool (often at night). – This is best for environments with temperatures below 55 degrees Fahrenheit for 3000 or more hours a year. – Using economizers, chilled-water-plant energy consumption can be cut by up to 75%. – You will also see reductions in maintenance costs, because the fluid-chilled cooling system allows you to drastically reduce—maybe even completely eliminate—the need for chiller operation. – They not only save costs, but they don’t allow contaminants or altered humidity levels into the datacenter. – Water-side economizers work with a cooling tower, evaporative cooler, or dry cooler to cool down the datacenter. – This type of economizer is normally incorporated into a chilled water or glycol-based cooling system. – Fluid in the cooling system passes through a coil to cool the room, thus eliminating the need for the compressor to operate.  On-Demand Cooling These units are brought in to provide temporary cooling when central air is down. There are two types of on-demand cooling systems, very similar in function to economizers: Air to air : – Smaller air-to-air coolers can be wheeled into the room needing cooling. – They use flexible ductwork to connect to a window, and then the generated heat is transferred out of the building. – They can be plugged into a standard 110-volt wall outlet. – Larger units can be mounted on the outside of the building, with cool air being ducted through a window. – These units operate on temporary 208-to-230-volt circuits.  Water based : – These are much larger units, where a standard garden hose is connected to the device so that water flows in, cools down the equipment, and then is sent through a second hose to run down a drain.  HP’s Solution Hewlett-Packard offers a cooling technology that it says can cut an IT department’s power costs by upto 40%. The system, called Dynamic Smart Cooling, uses sensors to control the temperature in specific areas of the datacenter. HP labs were able to reduce the power to cool a datacenter from 45.8 kW using a standard industry setup to 13.5 kW  Dynamic Smart Cooling is an intelligent solution, and rather than turning your datacenter into a meat locker, the system allows air conditioners—managed by specially designed software—to regulate the cold air delivered to a room based on the needs of specific computers.  Dynamic Smart Cooling uses the datacenter’s air conditioning system to adapt to changing workloads with sensors attached to the computers. If the system senses that a computer is warming up too much, air conditioners will send more cool air.  Optimizing Airflow To deliver the precise cooling environment, air must be exchanged at a sufficient rate. Normal office environments must change air over twice an hour. In high-density datacenters, air has to be exchanged 50 times an hour. If enough air is not exchanged, cooling air will heat up before it reaches the equipment, and disaster could occur.  Good practices can help minimize your costs without you having to buy the newest product. Some best practices that can help optimize the airflow around your servers and other networking equipments are:  Hot Aisle/Cold Aisle: – Equipment is typically designed to draw in air from the front and then blow the exhaust out the rear. – The cool sides of equipment are arranged together, whereas the hot sides of equipment face each other. – This allows the equipment to draw in cool air, rather than air that has already been preheated by the rack of equipment in front of it. – The cold aisles have perforated floor tiles to draw cooler air from the raised floor. – Floor mounted cooling is placed at the end of hot aisles, but not parallel to the row of racks. – This is because parallel placement can cause the hot exhaust to be drawn across the top of the racks and mixed with the cool air. – It also decreases overall energy efficiency.  Raised Floors: – Datacenters are conventionally built on a floor that is raised 18 to 36 inches. – The higher the floor level, the more air that can be distributed under the floor and the more air that can be used by the cooling system. – But higher isn’t always practical. – There can be major disruptions to day-to-day operations. – Plus, the higher up you build the floor, obviously, the closer you’ll be getting to the ceiling. – This can be a hindrance not only for rack sizes, but also for the flow of air over the top of equipment.  Cable Management: – Developing a good cable management system in conjunction with the hot-aisle/cold-aisle design can equate to more energy efficiency. – Whenever possible, it’s best to route your cables under the hot aisle, as shown in Figure 4-5. – This reduces the cool air’s path to the equipment as it is drawn in through the perforated tiles and into the equipment's cooling systems. – Some racks now provide expansion channels that help with cable management and ease heat removal for high-density racks. – Some organizations are also running cabling above or through racks, rather than under the floors, to reduce the interference with the flow of air from below. – Some are deploying advanced power strips to bring the power load closer to the rack rather than running so many cables through the datacenter.  Vapor Seal: – It’s also important to ensure you have a good vapor barrier in your datacenter, cutting it off from the rest of the building. – If you have a poor vapor barrier, humidity will move into the datacenter during hot months and escape during the winter months. – A good vapor seal reduces the costs to humidify or dehumidify.  Prevent Recirculation of Equipment Exhaust Your networking gear can get hot enough on its own and doesn’t need help from its neighbors—nor does it need to heat up its neighbors. The following are some simple steps you can employ in your datacenter to prevent exhaust from being reabsorbed by other devices. 1. Hot-aisle/cool aisle: Employ the hot-aisle/cool-aisle design mentioned earlier. 2. Rigid enclosures: Build rigid enclosures to keep exhaust heat from being sucked back into the device’s cool air intakes. 3. Flexible strip curtains: Use flexible strip curtains to block the open air above your racks that have been configured into a hot- aisle/cool-aisle layout. 4. Block unused rack locations with blanks: Equipment typically draws in cool air from the front and exhausts it out the back. Blanking open areas under equipment prevents the exhaust from being drawn back into the device. 5. Design with cooling in mind : Although most do, some equipment does not draw air in from the front and exhaust it out the back. Some have top-discharge or side to- side designs. Configure your racks to ensure your equipment doesn’t blow into the intake of other equipment. 6. Select racks with good airflow: Buy racks that don’t have an internal structure that would block the smooth flow of air to your equipment. Supply Air Directly to Heat Sources Rather than shelling out the money to cool the entire datacenter, save some money and just cool down the devices generating heat. These tips can help: 1. Use the correct diffusers: The type of diffuser you would use in an office is not appropriate for a datacenter. Select diffusers that deliver air directly to the equipment that needs cooling. 2. Correctly place supply and returns: Diffusers should be placed right by the equipment to be cooled. They should not be placed so they direct cooling air at heat exhausts, but rather into the air intakes. Supplies and slotted floor tiles should not be placed near returns to prevent a cool air “short circuit.” 3. Minimize air leaks: Systems that use a raised floor can lose cool air through cable accesses in hot aisles. 4. Optimize air conditioner placement: In large datacenters, a computational fluid dynamics (CFD) model would be useful. This helps locate the best placement for cooling units. It also helps minimize the distance between air conditioner units and large loads. 5. Use properly sized plenums: Return plenums need to be the right size to allow a lot of air to flow through. Obstructions such as piping, cabling trays, and electrical conduits need to be taken into consideration when plenum space is calculated. 6. Provide enough supply: Under-floor supply plenums must be big enough to allow enough air to service your equipment. Again, take into consideration obstacles such as piping, cabling trays, and electrical conduits.  Fans Fans also suck up a lot of power, especially when a lot of them are spinning at the same time. Take these tips into consideration to improve fan efficiency: 1. Use a low-pressure drop system : Use low- pressure drop air handlers and ductwork. Make sure there is enough capacity in your under-floor plenums to allow air to flow. 2. Use redundant air handlers during normal operations: It is more efficient to use auxiliary fans at a lower speed than a single fan at high speed. Power usage drops with the square of the velocity. As such, operating two fans at 50% capacity uses less power than one fan at full capacity.  Humidity Datacenter cooling systems must also be able to adapt to exterior temperature and humidity. Because these factors will change depending on where on the globe the datacenter is located— along with the time of year—datacenter air- conditioning systems must be able to adapt to these sorts of changes. Too much humidity or too little humidity can wreck your datacenter equipment. Use these tips to help keep your datacenter at the right level: 1. Establish a humidity sensor calibration schedule: Humidity sensors drift and require frequent calibration—more so than temperature sensors. Also, incorrect humidity sensors are less likely to be noticed than incorrect temperature sensors. As such, establish a frequent test and calibration schedule for your humidity sensors. 2. Allow for sensor redundancy: Make sure you have enough sensors to keep an eye on your datacenter’s humidity level. To ensure a tight control, multiple sensors should be used. At the very least use two, but more are better. 3. Manage humidity with a dedicated unit: If ventilated air is used (maybe from an air-side economizer), control humidity with a single ventilation air handler. 4. Lock out economizers when necessary: When using an air-side economizer, minimize the amount of air that’s brought in when the dew point is low. This saves money on having to humidify the dry air. 5. Centralize humidity control : Each datacenter should have its own centralized humidity control system. Multiple systems wind up fighting each other, and the system becomes less efficient.  Adding Cooling If your datacenter is especially “equipment dense,” you’ll need to add some extra cooling capacity. The best way to cool your equipment is to make sure the cooling gear is as close as possible to the heat sources. When you decide how to supplement your cooling systems, you should consider what type of system to use (air or fluid based) and what type of design the system will use.  Fluid Considerations As anyone with a car knows, fluid is a great way to move heat from equipment (in this case, the engine) to keep it cool. Fluid-based cooling systems have to be used with care. Water isn’t the only fluid used for cooling. Though water is normally used in floor-mounted cooling, because of safety concerns, R134 a refrigerant is typically used when cooling is used closer to the equipment. This is because refrigerant turns into a gas when it reaches the air, so leakage doesn’t pose a threat to your equipment. Table 4-3 lists the advantages and disadvantages of both solutions.  However, it isn’t just safety and effectiveness that makes refrigerant a good match for cooling needs. Fluid solutions employ micro channel coils for better efficiency, and a low-pressure system results in lower operating costs. It can also provide an energy efficiency savings of between 25 and 35% based on kilowatts of cooling capacity per kW of heat load.  System Design Because getting close to the heat source is so important, the cooling system’s design is important to consider. There are two common designs in datacenters—open and closed. In a closed design, the electronics and cooling equipment are situated together in a sealed environment. The benefit of this is that it is a high-capacity cooling solution. The downside is that the design isn’t as flexible, nor fault-tolerant.  In a datacenter environment, however, an open design is preferred, because a closed solution offers little flexibility. For example, if a cooling system fails, the racks are isolated from the room’s own cooling opportunities. Inside the enclosure, the server can reach its over- temperature limit in 15 seconds. With an open architecture, modules can be positioned close to the racks, but are not enclosed, so room air can be a sort of backup if the cooling equipment fails. This makes it much safer for both your organization’s data reliability as well as the hardware’s physical health. Not least of all, you have much greater flexibility to configure and reconfigure your datacenter as the system evolves.  Datacenter Design You can optimize your cooling needs by how you design your datacenter. A number of issues can help you reduce the amount of cooling you need, simply by how you design your datacenter and how cooling is deployed.  Centralized Control When designing your cooling plan, it’s best to employ a custom centralized air-handling system. This sort of system offers several benefits over the prevalent multiple-distributed unit system, including the following: – Better efficiency. – Can use surplus and redundant capacity. – Units can work in conjunction with each other, rather than fighting against one another. – Uses fluid-cooled chiller plants, which are much more efficient than water- and air cooled datacenters. – Less maintenance is required.  Design for Your Needs Unfortunately, our datacenters’ power needs rarely get the exact fit they need. They are usually loaded too light. Although a certain amount of the dark arts are involved in getting the size right, it is important to get as close as you can with electrical and mechanical systems so that they still operate properly when under loaded, but are still scalable for larger loads. You can come close to this Zen-like balance if you consider a few issues:  Upsize the duct, plenum, and piping infrastructure. This reduces operating costs and allows a measure of future-proofing. Use variable-speed motor drives on chillers, chilled and condenser water pumps. Also, use cooling tower fans to help with part-load performance. This can be especially helpful when controlled as part of a coordinated cooling system. Examine efficient design techniques, such as medium-temperature cooling loops and fluid-side economizers. Cooling-tower energy use is typically a small portion of energy consumption. If you upsize cooling towers, you can improve chiller performance and fluid-side economizers. Although this involves a larger cost up front and a larger physical footprint, you’ll find savings in operational costs.  Put Everything Together Efficient cooling isn’t just a matter of installing intelligent equipment. Organization-wide considerations must be implemented, including design and decision- making issues. Such issues include: – Use life cycle cost analysis as part of your decision- making process. – Involve all key stakeholders to keep the team together on the project. Document and clarify the reasons for key design decisions. – Set quantifiable goals based on best practices. – Introduce energy optimization as early as possible in the design phase to keep the project focused and to keep costs minimized. – Include integrated monitoring, measuring, and controls in facility design. – Examine and benchmark existing facilities and then track your performance. Look back over the data and look for any opportunities to improve performance. – Evaluate the potential for onsite power generation. – Make sure all members of the facility-operations staff get site-specific training, including the identification and proper operation of energy-efficiency features.  As anyone who has tried to string network cabling through an old building knows, planning for the future can save a lot of time, energy, and money. The same philosophy is true of cooling systems—it’s a good idea to plan for the future. Selecting a technology that can scale to future needs is a critical part of your considerations. Because if you do add more power to your system, you’ll need to add more cooling. Most server manufacturers are working on solutions that bring refrigerant-based cooling modulated into the rack to manage heat densities of 30 kW and higher. This will make refrigerant-based systems compatible with the next generation of cooling strategies.  Your cooling system is such a huge portion of your datacenter that it really merits a lot of your attention—not just to reduce your electricity bill, but also to mitigate your carbon emissions. Spending time and effort to ensure you have a well-running cooling system will help not only your organization but also the environment. It isn’t just your machinery that can help reduce your impact on the environment. End