Data Centers from Edge to Cloud
Providing infrastructure solutions specializing in highly available, scalable, redundant and manageable, turnkey solutions for IT environments from utility-to-desktop
Cooling Systems Design Goals
To establish an effective cooling solution for any new or upgraded data center or computer room, it is essential to establish a set of design goals. Experience suggests these goals can be categorized as follows:
- Plan for increasing critical load power densities
- Utilize standard, modular cooling system components to speed changes
- Allow for increasing cooling capacity without load impact
- Provide for cooling distribution improvements without load impact
- Minimize the possibility for human error by using modular components
- Provide as much cooling system redundancy as budget will allow
- Eliminate air mixing by providing supply (cold air) and return (hot air) separation to maximize cooling efficiency
- Eliminate bypass air flow to maximize effective cooling capacity
- Minimize the possibility of fluid leaks within the computer room area as well as deploy a detection system
- Minimize vertical temperature gradients at the inlet of critical equipment
- Control humidity to avoid static electricity build up and mold growth
- Deploy the simplest effective solution to minimize the technical expertise needed to assess, operate, and service the system
- Utilize standard, modular cooling system components to improve serviceability
- Assure system can be serviced under a single service contract
- Provide accurate and concise cooling performance data in the format of the overall management platform
- Provide local and remote system monitoring access capabilities
- Optimize capital investment by matching the cooling requirements with the installed redundant capacity and plan for scalability
- Simplify the ease of deployment to reduce unrecoverable labor costs
- Utilize standard, modular, cooling system components to lower service contract costs
- Provide redundant cooling capacity and air distribution in the smallest feasible footprint
Once your design goals are established there are a number of additional steps recommended for data center cooling best practices.
Determine the Critical Load and Heat Load. Determining the critical heat load starts with the identification of the equipment to be deployed within the space. However, this is only part of the entire heat load of the environment. Additionally, the lighting, people, and heat conducted from the surrounding spaces will also contribute to the overall heat load. As a very general rule-of-thumb, consider no less than 1-ton (12,000 BTU/Hr / 3,516 watts) per 400 square-feet of IT equipment floor space.
Establish Power Requirements on a per RLU Basis. Power density is best defined in terms of rack or cabinet foot print area since all manufacturers produce cabinets of generally the same size. A definite Rack Location Unit (RLU) trend is that average RLU power densities are increasing every year. The reality is that a computer room usually deploys a mix of varying RLU power densities throughout its overall area. The trick is to provide predictable cooling for these varying RLU densities by using the average RLU density as a basis of the design while at the same time providing adequate room cooling for the peak RLU and non-RLU loads.
Determine the CFM Requirements for each RLU. Effective cooling is accomplished by providing both the proper temperature and an adequate quantity of air to the load. As temperature goes, the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) standard is to deliver air between the temperatures of 68 °F and 75 °F to the inlet of the IT infrastructure. Although electronics performs better at colder temperatures it is not wise to deliver lower air temperatures due to the threat of reaching the condensate point on equipment surfaces. Regarding air volume, a load component requires 160 cubic feet per minute (CFM) per 1 kW of electrical load. Therefore, a 5,000-watt 1U server cabinet requires 800 CFM.
Perform Computational Fluid Dynamic (CFD) Modeling. CFD modeling can be performed for the under floor air area as well as the area above the floor. CFD modeling the airflow in a computer room provides information to make informed decisions about where to place CRAC equipment, IT-equipment, perforated tiles, high density RLUs, etc. Much of the software available today also allows mapping of both under floor and overhead airflow obstructions to more accurately represent the environment.
Determine the Room Power Distribution Strategy. The two (2) main decisions in developing a room power distribution strategy are: (1) Where to place the power distribution units (PDUs)?, (2) Whether to run power cables overhead or under the floor?
Determine the Cabinet Power Distribution Strategy. In deciding how power will be distributed through the cabinet, use of dual power supplies, and cabling approach, it is important to understand the impact of power distribution on cooling, particularly as it is related to air flow within the cabinet.
Determine the Room & Cabinet Data Cabling Distribution Impact. Typically, there are three (3) choices in delivering network connectivity to an RLU. They are: (1) Home run every data port from a network core switch, (2) Provide matching port-density patch panels at both the RLU and the core switch with pre-cabled cross-connections between them, such that server connections can be made with only patch cables at both ends, (3) Provide an edge switch at every rack, row, or pod depending on bandwidth requirements. This approach is referred to as zone switching.
Establish a Cooling Zone Strategy. Recall that effective computer room cooling is as much about removing heat as it is about adding cold. Generally speaking, the three (3) equipment cooling methods along with their typical cooling potential can be determined from the following table:
Room Cooling ~2 kW per RLU Row Cooling ~8 kW per RLU Cabinet Cooling ~20 kW per RLU
It is also critical to consider high-density cooling and zone cooling requirements.
Determine the Cooling Methodology. Upon determining what cooling zone will be required, the decision of what types of air conditioners will be needed, must be made. There are four (4) air conditioner types: (1) air cooled, (2) glycol cooled, (3) condenser water cooled, (4) chilled water.
In addition, it is also important to determine how heat will be rejected within the system and what type of cooling redundancy is required and available for a particular methodology.
Determine the Cooling Delivery Methodology. Different architectural attributes affect cooling performance in different ways. For instance, designs should consider the location of the computer room within the facility (I.e. onside versus inside rooms), height of the raised floor, height of suspended ceiling, etc.
Determine the Floor Plan. The ‘hot aisle / cold aisle’ approach is the accepted layout standard for RLUs for good reason. It works. It was developed by, Dr. Robert Sullivan, while working for IBM and it should be adapted for both new and retrofit projects. After determining the hot/cold aisles it is critical to place the CRAC units for peak performance. This may include room, row, or rack based cooling approaches. Each works well depending upon the IT infrastructure, power densities, CFM requirements, and other attributes previously discussed.
Establish Cooling Performance Monitoring. It is vital to develop and deploy an environmental monitoring system capable of monitoring each room, row, and cabinet cooling zone. A given is that once effective cooling performance is established for a particular load profile, it will change rapidly. It is important to compile trending data for all environmental parameters for the site such that moves, adds, and changes can be executed quickly.
Contact PTS for your Next Data Center Project
You shouldn’t have to compromise when it comes to your data center. Our expert team assesses your unique needs, and then employs a proven data center plan that reduces the amount of support infrastructure you will need.
The result is higher efficiency, reduced complexity and better resiliency at lower cost… all at the same time.
Contact us for a quick chat to discuss how a smarter design for your data center can save you money.
Contact PTS today:
Pete Sacco (Founder and President, PTS Data Center Solutions) and Jamie Funk (Head of Sales for Verticals at Bloom Energy) talk to Dan Loosemore of Data Center Dynamics about future-proofing the modern data center using PTS’ Data Center Reimagined approach and Bloom Energy’s fuel-cell and micro-grid solutions.
Webinar Data Center Development Innovation for the New Age of Facilities. Learn how end users and providers are determining cloud or on-prem strategies for new facilities coming online, how fully integrated edge developments have changed the need for on-prem solutions, who are changing their migration to the cloud and where the opportunities are for expansions.
The average enterprise data center costs between $10 million and $12 million per megawatt to build, with costs typically front-loaded onto the first few megawatts of deployment. Seasoned data center builder Peter Sacco, founder and CEO of PTS Data Center Solutions, claims that it is possible to drive down costs by around 25 per cent or more, simply by being more disciplined on design.
How can data center operators insulate themselves from the increasing risks posed by aging centralized power grids prone to outages and vulnerable to natural
It’s vital for companies to invest in data center operations that are simple in design, easy to operate, and minimize infrastructure needed to achieve
Intel’s IT Chief Technology Officer Shesha Krishnapura sat down for a Q&A with Data Center Dynamic’s Kisandka Moses to discuss Intel’s cost-efficient data center
There are many factors influencing how suitable a given data center is at meeting the needs of the organization that is utilizing it. Inevitably,
With the size of stored data increasing at an astronomical rate, there’s no question that the data center market is booming – no matter whether it’s an on-site data