The Data Center Blog

Proper initial planning of data center facilities.

When initially tasked with creating a data center environment the most basic constraint, the one that will form the foundation of your solution is to understand the relationship between available space, power and cooling.. While there are different approaches as to deciding which side of the triangle is the most important, there is universal consensus that these are most important design considerations.

Before the deployment of blade server technologies, average power consumption per rack was typically 5-8 kilowatts (14 Compaq Proliant DL380 G1 Servers with two 275W power supplies = 7700 watts). With the advent of blade server implementations, power requirements skyrocketed to up to over 20 kilowatts per rack (6 Compaq C-Class Blade Enclosures can require 3255 watts each = 19530 watts). Couple this with the price of electricity steadily increasing an additional $0.04 - $0.14 per kilowatt hour within the past five years and you can see why power is a critical design element for any data center.

In complement with the calculations of kilowatt power per rack in your data center, there must be proper cooling. There are two specific factors to consider when sizing cooling requirements; first the actual amount of BTUs (British Thermal Units) required to dissipate the generated heat, and equally as important is the amount of airflow which is measured in CFM (cubic feet per minute). Unlike determining power requirements, which can be easily figured out as multiplying the number of power supplies with the watts generated from these supplies, calculating the amount of BTUs can be a bit more complex. A good rule of thumb in determining the amount of heat generated per rack is to take 30% of the total wattage and assign that as the heat cost. So to use the above example of the 6 C-Class Blade Enclosures 19530 watts generates 5859 watts of heat. To convert this into BTU’s, multiply the kilowatts by 3412.14. (5.859 KW * 3412.14 = 19,991 BTU), or this can be expressed in cooling tons by dividing the BTUs by 12,000 (19,991 BTU / 12,000 = 1.66 Tons). The second step is to determine the CFM of cooling.  Legacy chiller solutions cool the entire room and as such are the least efficient ways to go as you are applying energy to cool the room, and not the equipment inside of it. Today’s chillers reside either inside of the rack or off to the side of it, or you will find situations of an enclosed area made up of the racks creating hot aisles and cold aisles. This is a more efficient solution as you are directing the chilled air directly on the equipment and not the ambient air in the room. To accurately measure how many CFM per minute that you need in your rack, it is a bit more complex, in fact you may wish that you stayed awake in high school physics. First, you need to determine the maximum “hot” temperature – meaning air leaving the rack, for example (30°C/86°F). Next determine the “cold” temperature – meaning the air going into the rack, for example (15°C/59°F). Lastly, you need to calculate the total amount of heat generated in kilowatts (5.859 kW). Calculate the amount of air that is needed to be moved using the specific hear of air (1 kJ / kg K). 5kJ/sec means that you will need 5kg/sec for a 1 degree rise, however in this case, with a 15 degree rise this converts to .33kg /sec. To turn .33kg/sec into CFM you multiply .33kg to 1.3 m3/kg for a total of .43m3/sec. There are 35 cubic feet in a cubic meter, so you need about 15 cubic feet per second, or about 900 CFM.

After two segments of the triangle have been calculated, usable space, and its relationship can be examined. Referring back to the Identification of the components of a data center, there are specific components to plan for in a data center design: The room that the server racks reside in, room for the power and mechanical services, the communications demarcation point, and the operations command room. The dimensions of these areas are dependent on the requirements of the data within the data center but after this initial allocation, the remainder of the physical location is determined to be your “available space”, which is measured in rack spaces or 3’x4’

When combined together, one can quickly understand how the combined factors of power, cooling and space form the most important factors in designing a successful data center.

Two quick items of note coming from Cisco.

I virtually attended Cisco’s C-Scapethis week, and it was mind blowing (as always) to see what and where Cisco is positioning themselves. If you get the chance, catch up on the video sections.  In the interests of full disclosure, I am a Cisco FanBoy, and was also a speaker at the GSM this past summer, but none the less, the information that they presented is indeed quite good. One of the side discussions that was brought up was Cisco’s concept of Data Center 3.0 which is interesting to compare to Microsoft’s Generation 4 (which has been discussed previously on this blog). Today, Cisco released a very informative description of what specifically, they mean to accomplish with Data Center 3.0.  Watch the video of it here. (17 minutes)

So then news comes today with some more details on Cisco’s move into the data center, and all I can say is… WOW!

Expected next year are internally developed datacenter blade servers, energy efficiency improvements across Cisco’s switching portfolio, and a new release of the company’s unified communications software for intercompany collaboration.

The product launches are intended to buttress Cisco’s strategy to become not just the leading network vendor to corporations and services providers, but to become the leading supplier of overall IT architectures to these constituents.

“The network will enable all forms of communication and IT,” said Cisco CEO John Chambers during his keynote address at the company’s annual C-Scape analyst conference here last week. “IT is not enabling our strategy, it is our strategy.”

From Infoworld

Similar to the PMBOK project management triangle, I theorize that there are similar triangles that come into play during the design of a data center.

Referencing the project management triangle, the success of a project is based on three constraints, Scope, Cost, and Schedule. Throughout the design and planning of a data center, there are three other versions of the triangles:

·         Space, Power, and Cooling

·         Service, Cabling, and Network Bandwidth

·         Information, Backup, and Recovery

The first triangle deals with the physical layout and constraints of a data center. There is a direct correlation between usable space and power and cooling requirements. For example, if power and cooling requirements are a premium, they will affect the usable space available to you. If usable space and power are constraints, that will limit the amount of cooling available to the equipment serviced inside of the data center. Conversely, if space and cooling are constraints, you must reduce the power available to the data center to face heat issues.

The second triangle is associated more with the connectivity to and within a data center. As with the first triangle, there is a direct correlation between a service level, the cabling plant infrastructure and available network bandwidth. If a high service level is the main constraint, you must either increase the cabling (and ports) to provide sufficient connection, or increase available bandwidth to ensure a minimum level of always available service. Increasing bandwidth can reduce cabling and increase the level of service available to the equipment residing in the data center. If the available cabling plant is a constraint, this will affect the available network bandwidth and impact the level of service available to the equipment.

Lastly, in order to properly plan for the equipment that is necessary in the data center, one must identify the data that is residing within the data center and understand the impact that it will have on the logistics of backup and recovery (both in regards to time, and space). Small recovery windows must drive either the amount of data or types of applications that can be serviced inside of the data center or the medium by which this data is backed up. Fixed data requirements, will either constrain the recovery time or the backup medium. While a best of class backup solution can increase the available data size and possibly decrease recovery time.

It is the goal of this continuing series to delve deeper into each of these triangles, spending time examining and expanding upon them. As they are examined, I would appreciate any feedback.

Yesterday, we reported on Dell’s entry into the data center-in-a-box market with their container based solution. Today comes news, and a video from HP and their entry in the data center-in-a-box market with their POD (Portable Optimized Data Center). Datacenterknowledge.com has put up a video tour of the POD by Steve Cummings. Some of the differentials between HP and Dell’s solutions: one pod vs. Dell’s two shipping container, 50U racks that fill the entire interior, making for a solid barrier from hot air, and get this… the POD operates at peak efficiency when the cold aisle is at 90 degrees!!!!

Watch the video here:

The HP POD

People can often confuse what is termed as a “computer room” or “equipment closet” with a facility designed as a data center.

A computer or equipment room is just a secured location that houses a company’s computer equipment, and is the most basic of all designs that many small businesses or operations employ to maintain their operations. More often than not, it is too small, usually a small room or closet that was not originally purposed as a computer room, and doesn’t have adequate environmental capabilities.

A properly designed data center is a culmination of four distinct layers of spaces:

·         A computer room – where your computer equipment physically resides.

·         The data center itself built around the computer room that includes a dedicated entrance both for power, data and mechanical services but also for human access , an isolated mechanical and electrical room, a central telecom room that acts as the central demarcation point for your telecommunications, an operations center that monitors all of the activity of the computer room and the data center, offices for the staff, and lastly a method of receiving and storing large or oversized equipment such as server racks, batteries and alike.

·         Since a data center will often reside within a company’s operation, or business location, the building itself has a level of importance as well. While its components may be the same as components found in the data center it is important to note that the in order to provide a secure design, those components must remain separate from the equipment powering the data center. This will include the equipment rooms that support the building itself, and the office space for general employees, and the generic “business area” where your company performs its business.

·         All of this is surrounded by the building site itself. As the largest and most prominent layer, the location of the building must be secure and in good working order. Careful attention to its geographical placement in regards to environmental concerns (tornados or hurricanes for example) must be balanced against a reliable level of utility services and transportation corridors.

One of the most critical steps in deploying and designing data centers is with planning. Understanding how a data center will operate in conjunction with the business model  will go a long way in determining the proper design, scaling, and staffing of your solution.

According to a recent article in the New York Times, Dell is exploring the idea of a “Data Center in a Box” market to provide a solution for companies that don’t want to spend the time and money to build out a data center, or in cases of the military, provide an on the fly data center that can be easily deployed - just hook up your utlities and off you go.

Who is the big driver of this?

That’s easy: Microsoft.

Microsoft has been the main advocate of containers, saying they will form the basis of its future data center designs. Some of Dell’s first containers will go to a new Microsoft data center near Chicago, according to Forrest Norrod, the vice president in charge of Dell’s Data Center Solutions business.

And Microsoft’s interest in the container idea should inspire others to take a look at the technology.

“I think next year will be the year for this,” Mr. Norrod said.

Whereas competitors have put all of the requisite technology components into a single container, Dell has gone with the double-decker idea. One container is full of server, storage and networking systems, while another container handles power and cooling. By using this design, Dell claims it can stick with standard equipment across the board, saving customers money and making it easier to upgrade the units.

Each set of containers holds about 1,300 servers and consumes about as much power as the homes making up a suburban subdivision. The cost can easily top $500,000.

Data Center in a Box, by Dell

Created in 2005, The Data Center Standard defines the specifications for data center infrastructure. Infrastructure is defined in the following manner:

·         Site space and layout

·         Cabling infrastructure

·         Tiered reliability

·         Environmental considerations

Additionally, the standard was created to provide a standard for technical professionals to work within during the planning and development process of data center design and implementations. Before its release, there was no uniformed standard by which to measure such facilities and designs against and as such, different solutions were released and to the uneducated consumer, or professional the entire concept of data center, what it actually is and what it is actually supposed to do became muddied.  The 942 standard solves this.

Unlike many of the other standards published with by the TIA, and contrary to initial impression, the majority of the 942 standard is devoted to the specifications of the facility housing the data center, and not the computers or systems that reside in it. Considerable time is spent on the cabling / telecommunications topology of the data center, specifically on the architecture and the media of the structured cabling systems, and the accommodation of a varied range of networking applications and protocols.

The TIA-942 standard also provides for a clear measurement for a tiered standard which is recognized universally. Before the release of the standard, different organizations could release different standards confusing the general audience, and not allow for a true apples to apples comparison of existing data center design, and an effective way to present data center design requirements.

Environmental considerations such as fire suppression, temperature and humidity control, structural design, electrical and mechanical operation are also covered in the TIA-942 standard. Depending on the tiered reliability of the datacenter, the environmental requirement can change, however there is a minimum baseline standard in regards to environmental standards that must be adhered to.

In summation, as the concept of the data center grows and permeates through the operations of many businesses and more and more technology professionals are tasked with providing “can’t fail” solutions for their organizations, the TIA-942 standard establishes a clear, universally recognized standard to help in the planning, preparation and design of the modern data center environment.

For more information on the TIA-942 Standard, visit the TIA at www.tiaonline.org/standards.