A study of factors affecting data center airflow in a closed aisle.
Written by Rajesh Nair, Founder & CTO of AdaptivCool, an EEC partner
What happens to the data center airflow through a high-density server when the cold aisle it is in is capped or contained?
Of course, no one would let us do this experiment in their critical data center. So we had to resort to simulating the effect on a computer using a CFD tool.
Every server in a rack is designed to pull enough air through it to cool the electronics in it under all specified environmental conditions. The front and rear doors of a rack and cabling in it can add additional series flow impedance to the internal impedance of the server. The server fans can overcome additional impedance to some extent since they are normally over-designed. Having enough cold air available in an open cold aisle helps overcome some of these airflow restrictions. However, when the cold aisle is capped (contained) the situation changes a lot.
In the simulation study below, we modeled an existing data center. The pressure pattern along the whole data center cold aisle is shown in Figure 1 as shaded plane. The pressure stays relatively flat along the whole length of the room. We modified the model by capping the cold aisle and studied the pressure in the same plane. Nothing else in the room was changed. As the figure shows, the pressure in the cold aisle dropped significantly. A lower server intake pressure increases the pressure difference the server fan has to work against, causing the airflow to drop. Here is an electrical engineer’s analysis of aisle capping on server pressure and airflow.
Airflow through a server enclosure:
Server boxes are designed to operate under wide thermal environmental conditions. The built-in fan is capable of pulling enough air to cool the electronics in the box. In Figure 2, a simple electrical representation of airflow is shown. The air intake and exhaust are the open room space and are at the same pressure. The fan creates enough pressure (Ps) to create airflow (Q) over the internal impedance (Zs). The fan speed determines the driving pressure and airflow. The internal control system modulates the fan speed to draw just enough air to cool the electronics. Once this server is placed in a rack it sees additional impedance, Zr, from the front and rear doors and cabling in the path of airflow (Figure 3). The server fan now needs to operate at a higher speed to compensate for these series impedances to get enough airflow to cool the devices inside. This may be acceptable to a limit since these systems are normally over-designed. But at the extreme thermal conditions this additional impedance, Zr, may limit the cooling capacity of these servers. So it is important to minimize Zr.
Airflow in a closed aisle:
Once the cold aisle is closed, the airflow-pressure scenario changes further. Since the aisle is closed, the aisle pressure (Pa) may not be at the same as the pressure in the hot aisle. As shown in Figure 4, there are two series airmovers in the airflow path; the CRAC and the server fan. The pressure created by the CRAC (Pc) causes airflow under the floor to overcome the under floor impedance (Zu) and the impedance of tiles in the cold aisle (Zt). Depending on the distance from the CRAC to the aisle and underfloor impedance, the airflow could vary significantly from one closed aisle to another. In Figure 1 the CRAC fans were not creating enough pressure to discharge air into the closed aisle. In fact, the fans in the servers were pulling cold air into the aisle thus causing a lower pressure in the aisle than the rest of the room. The net airflow through the servers dropped significantly once the aisle is closed. One fix to increase airflow through the servers is to increase the aisle pressure by using more open grates (and thus reduce Zt). This could still only be a temporary solution. The pressure in an aisle is influenced by multiple CRACs. If the CRAC with highest influence is turned off, the aisle pressure can immediately drop and other CRACs may not be able to compensate.
Pressure compensation in a closed aisle:
The way to guarantee airflow through servers in a contained aisle is by controlling the pressure in the aisle. An effective way is by use of either a fan assisted active tile (Demand Based Cooling system from AdaptivCool), or fan assisted chimney or rack backdoor with automatic speed control. The last two options need to be installed on each rack, as these manage pressure drop only within a rack. The active tiles (Figure 5) can control pressure in the whole aisle and compensate for CRAC failure. The cold to hot aisle pressure difference is used for fan speed control. This approach can offer normal server operation in a closed aisle even under CRAC failure conditions. Figure 6 is a representation of airflow and pressure in the room.
Closing:
The primary advantage of data center aisle containment is improved cooling efficiency, by avoiding mixing of cold and hot air in a data center. It also comes with unintended consequences. The two important ones are decreased data center airflow and loss of redundancy to CRAC failures. Both of these can be addressed by using additional fan assisted tiles in the airflow path that can hold a steady aisle pressure as well as compensate for CRAC failures.
To learn more about data center airflow best practices, fill out our contact us form or contact Brad Morgan, our Data Center Cooling Solutions Manager at bmorgan@eecnet.com or 508-229-1446.
During extraordinary events such as the Santa Ana Wind Storm in Southern California, Hurricane Irene and the rare October Snowstorm that hit the Northeast, EEC requires all personnel to be on-call and available to handle multiple emergencies for our critical data center and telecom customers. Take a look at what one of our data center customers had to say about our response during Hurricane Irene:
Access Northeast is a rapidly growing Data Center Provider with a focus on Colocation, Cloud, Hosting and WAN Management. We provide 24x7x365 mission critical services to our clients with an expectation of 100% application availability. As part of our emergency preparedness procedures in advance of Hurricane Irene in August 2011, we deployed additional staffing to our Marlborough, MA and Waterbury, CT data centers to monitor these facilities during the storm.
At approximately 1:00AM on Sunday, August 28, 2011 during the full brunt of Hurricane Irene, our Waterbury Data Center Manager, Douglas Barry, identified a leak at the condensate pump on one of our air handling units. He immediately called the Electronic Environments Corporation (EEC) dispatch center to request an emergency site visit. At 2:30 AM, EEC technician Steven Thibodeau arrived onsite. Steven was able to clear a blocked drain and returned the unit to full operation within the hour, thereby preventing any service impacting issues from transpiring that could have affected customer uptime.
Responding to an emergency maintenance call in the middle of the night is no easy task. Driving through the heavy rain and dangerous winds to reach a service site is also daunting. Resolving an HVAC unit issue in a timely manner is a tremendous benefit to any service provider. However, delivering all three in unison is the EXACT reason why Access Northeast has chosen EEC as our facility maintenance partner throughout New England. Due to the extraordinary efforts of EEC and Steven Thibodeau, Access Northeast was able to deliver uninterrupted service to our customers throughout the duration of Hurricane Irene, unlike many of our competitors in the region. Special thanks to all parties once again.
Sincerely,
Michael Donnellan,
Access Northeast Chief Operating Officer
Don't get left in the dark the next time there is a storm. EEC's 24x7 e
mergency repair service program is an aggressive and rapid response mechanism developed specifically for Data Center and Telecommunication operations. Our philosophy is to aggressively and rapidly respond to instances of system failure with the right personnel and the right material to bring your site back up immediately.
Our technicians arrive with "crash kits" containing replacement parts for 90% of failures and have access to a vast parts inventory - we will do whatever it takes to get your systems back on-line fast.
Join Electronic Environments and Packet Power for a free webinar:
Data Center Power Monitoring Made Easy
Recorded version now available > >
Power is an essential part of data center operations. Data centers must make sure enough power is always available. This has always been a challenge, but the challenge has grown as data center operators have to contend with delivering far more power per square foot of space -- and delivering it more efficiently.
Join us and you will walk away with great tips on how power monitoring can help you:
- Improve operational effectiveness
- Better allocate power costs
- Extend the life of your data center
You will also get:
- A comparison of data center power monitoring options
- Benefits of wireless power monitoring
- Simple ways to use monitoring data
Recorder version of webinar now available - click here > >
For more information on Electronic Environments or Packet Power and the soltuions and service we provide, please go to:
www.eecnet.com www.packetpower.com
The easiest, most flexible and cost-effective way to monitor power in your data center.
I just saw an interesting article on Computerworld.com "A tale of two U.S. government data center projects" where one project came in on-time and within budget and one went over budget and was late. The moral from the story is that the project that went over budget used the Design-Bid-Build (traditional bid/spec) approach and the one that came in on-time and within budget opted for a design/build approach – where you have a single point of contact and that contact is very familiar with the project from the start and can make any necessary changes fairly quick. Many times, as in the case of the over budget project, when a problem arises you have to negotiate with multiple teams to get things fixed which more often than not results in delays.
At EEC we work with our clients on either type of approach, however, we recommend the Design/Build approach. The need to bring technology projects on-line fast challenges traditional construction practices. The drawing below depicts the scheduling difference between a traditional construction schedule and a design/build schedule on a typical job:
The typical characteristics of the design/build approach are faster completion time (as represented here), the potential to be more cost effective, and turnkey responsibility is low risk for the client.
The data center construction phase is impacted by all the preceding events, missing even the smallest detail could wreak havoc later on. EEC has project managers 
that are thorough in making sure that all key milestones are being followed and are on track. The procurement phase is a critical component of construction management; our project manager devotes considerable effort and attention to the organization of this function. They establish key dates and monitor submittal and delivery cycles for long-lead equipment. This requires a keen eye on vendor management – developing an early, clear, and concise view of every vendor’s role.
Also, our internal network of resources including our product specialist, electrical and mechanical engineers, and specialized technicians all work together to make sure construction standards, schedule, and design specs are adhered too.
Have you recently built a data center or planning to build one? Which approach did/will you take? We’d like to hear from you.
There are a few important questions that facility and data center managers need to ask themselves to make sure they are on the right track in terms of timely and proper com
puter room air conditioning (CRAC) equipment maintenance; these include: What maintenance needs to be done on the equipment? When? How often?
Preventive maintenance is the best approach a company can take to limit the amount of downtime on equipment in critical environments. Within the CRAC system the most common failures tend to be high temperature or high head failures - both of which can be prevented by changing air filters frequently as well as by power washing the condenser to maximizer air flow and to maintain the operation of the unit.
I recently came across an article on SearchDataCenter.com, "Best practices for maintaining a data center CRAC unit". This article reviews several best practices that are important for maintaining your data center CRAC unit. Here are some of the items it covers:
-
Cooling is too critical to leave to chance
-
Fear of intentional cooling shutdown
-
What CRAC unit maintenance should entail
-
Taking care of mechanical items
-
The importance of refrigerant levels, electrical testing
-
Making time for external maintenance
Be sure to take a look at the article, it offers some great information in regards to maintaining your Computer Room Air Conditioning units.
And don't forget - performing maintenance can help you save energy
Perfo
rming maintenance can also help equipment run more efficiently, which in turn helps with the longevity of the equipment, but more importantly saves money by reducing the amount of energy the equipment uses. Scopes of work to help equipment run more efficiently should include regular checks of the quality of fluids, frequently changing air filters, and power washing condenser coils to name a few. Checking fluids is essential because contaminants can result in diminished performance. Changing filters frequently is important because it will maximize the airflow in the equipment. One of the most important maintenance checks on CRAC equipment is regularly cleaning the condenser coils. Condensers are known to draw in dirt, pollen, and other debris that restricts airflow and reduces efficiencies. According to the Consortium of Energy Efficiency, facilities in which proper CRAC preventive maintenance is completed will use approximately 15% to 20% less energy than those where systems are allowed to deteriorate. With the high cost of energy it is crucial to maintain the equipment in a data center in order to take advantage of efficiencies and save money.
Companies should work with their maintenance provider to develop a scope of work - with the exorbitant cost of downtime, a strong preventive maintenance program is not an option - it's a must!
Dirty fuel is a fact of life. Even with the development of cleaner burning fuels, contaminants are still a major concern when it comes to fuel systems. Micro-organisms can quickly become a major problem. Colonies of fungus and bacteria feed on your fuel.
Commonly known as algae, they spread rapidly in the presence of moisture. They easily spread from one tank to another by riding in the fuel. As these micro-organisms accumulate, they will spread through your data center's generator fuel system and quickly plug the fuel filter. Most end users will simply add a biocide to kill the fungus. While this does kill the bacteria, it does not remove the growth media from the tank. Biocides are also very toxic and cause chemical burns when they come in contact with the human skin.
The correct solution to the problem is to remove the water. This can be done by polishing the fuel in your generator fuel tank. Water is removed by three common methods:
- Stripping-which is a silicon based medium that inhibits the passage of water but allows fuel to flow.
- Coalescing-Gravity drags water droplets out of the slow-flowing fuel.
- Absorption-A filter medium with a high affinity for water and a low affinity for fuel absorbs the water.
- Algae-X in-line conditioner - by stabilizing and conditioning the fuel, reducing the size and mass of clusters of fuel molecules.
In addition to the above methods, particulate filtration is defined as passing a fluid through a filter media that has a specified micron rating. The fuel is allowed to pass and the particulate matter will be caught in the filter media. Bear in mind that the particulate matter that is smaller than the micron rating of the filter media will also be allowed to pass through the filter. Because of this, it is important that your primary filter be of a higher micron rating for the primary fuel pump's protection and that your secondary filter be of a smaller micron rating to protect the fuel injection pump and injectors.
What is your plan for keeping your data center up and running 24x7? Make sure fuel cleaning/polishing is in that plan. To learn more contact us.
Don't miss out on this great networking event designed to navigate telecom network operators through the many colocation and connectivity options in and around the NY Tri-State area. Telecom Exchange is designed to grow the neutral playing field and showcase all colo providers, data center providers and local and international network providers that are located in the NY area.
Electronic Environments will have a table at the show, stop by and say hello to Mike Farber and Ed Tarnawski - they will be able to help you with all of your telecom or data center physical infrastructure needs, including design/build, site repair & maintenance, cooling and energy efficiency solutions. EEC has been providing these services to telecom and data centers site for 25 years.
Telecom Exchange will be held on June 29th at the Cipriani Wall Street in NYC - be our guest and receive 20% off the cost of registration.
We look forward to seeing you there.
The Hanover Insurance Group was looking to improve their data center’s cooling efficiency, as well as the current system’s capacity – while allowing for increased IT (Information Technology) equipment densities within their data center at their headquarters in Worcester, MA.
EEC, working closely with The Hanover, recommended a unique solution that could not only save energy and eliminate hot spots, but also maximize IT cooling capacity in the data center. This system, Demand Based Cooling™ (DBC) airflow and thermal management system from AdaptivCool, consists of a series of networked HotSpotr™ underfloor and overhead air movers, server rack and Computer Room Air Conditioner (CRAC) temperature sensors, sensor hubs, CRAC controllers and Environmental Management Software.
The Challenge:
- Integrate new energy-efficiency technology within the data center, without affecting day-to-day operations
- Eliminate all the data center’s hot spots
- Achieve all this without making major changes to the existing data center infrastructure
- Meet the Company’s less-than-24-months economic payback criteria
Summary of Project:
- Cost of project: $144,000.00
- Estimated annual savings: 600,000 kWh
- Utility incentive: $21,500.00
- Company payback: 16 months (with rebate)
Benefits:
- Five out of 13 CRAC units were placed into hot-standby, and energy usage for cooling the data center was reduced by 27%
- Increased cooling redundancy
- Increased IT load capacity by 70kW
- IT equipment is thermally safe
- 24x7 monitoring allowing for automated CRAC unit response
- Energy incentive from the National Grid Commercial Energy Efficiency Program
To learn more about the project, the solution and other details involved:
I just saw this very interesting article on Information Week regarding a study of outages at U.S.-based data centers by the Ponemon Institute.
According to the study, data center failures cost an average of $5,600.00 per minutes. WOW - that is an incredible statistic. The article also goes on to say that respondents of this survey said the primary causes for unplanned outages include uninterruptible power supply (UPS) battery failure, water, heat, or cooling failure; and generator failure.
Check out the article at: http://www.informationweek.com/news/hardware/data_centers/229500121?cid=RSSfeed_IWK_All
Have you had any outages in your data center recently? If so, let us know what the primary cause for the outage was?
If they were caused by UPS battery, HVAC or generator failure - do you currently have regular preventive maintenance on these systems? Or what about battery trending or infrared thermography to find problems before they turn into an outage. If not, contact us - we can come out and do an assessment of your power and cooling to help you understand where you may run into problems in the future.
As we celebrate 25 years of providing mission critical services to the information technology and telecommunications markets, we would like to thank all of our customers, both past and current, who have helped us reach this significant milestone. We look forward to another 25 years of success and valuable relationships.
In honor of our 25th anniversary, Electronic Environments is offering the following incentives May 1st - December 31, 2011:
- Free additional three months when your purchase a 12 month full maintenance service contract (new contracts only)
- Free six months preventive maintenance on any newly purchased UPS, Generator or HVAC equipment (within 150 miles of any EEC regional service office)
- Free six month preventive maintenance inspection when you purchase a set of replacement batteries ($3000.00 minimum purchase)
- 25% off on the following services and products (For existing customers - discount applies if you are not already taking advantage of these services now):
- 10% off on the following services and products (For existing customers - discount applies if you are not already taking advantage of these services now):
**Note: Promotion ends on December 31, 2011 - all contracts and/or purchase agreements must be signed by by that date in order to receive any of the above incentives