Build Your Skills: The nuts and bolts of building a network rack

Get an overview of all the elements to take into consideration when you build your own server rack.

You've finally convinced your management that there is no more space in the server room or server closet for even one more tower server. It's time to move into the world of racks. So now what? Here are the fundamentals you need to know to make your first network rack successful.

Sizes and mounting
There is a common misconception that all rack cabinets are the same. They do adhere to the 19-inch rack mount standard, but this is a standard for width, not depth. The problem is that different racks have different depths, and, as a result, the devices that fit well in one rack may not fit in another.

Beyond rack cabinet depths there are some other basic measurements that you need to consider. Perhaps the most common mistake is thinking in terms of height in inches instead of in rack units, called U. One U is 1.75 inches in height. It is a standard measurement for most racks.

For instance, a full-size rack might have internal space for 42U worth of equipment—a little more than 6 feet. But the overall height of the rack, including the wheels and top, may exceed 7 feet.

In addition, the cables used in the rack will need to be longer than you might expect. Most rack mount servers include cable management systems. These cable management systems allow the server to slide out without disconnecting the cables. The problem with this type of system is that it consumes approximately three to four feet of extra cable length. While this is typically not an issue for network cables, it can be a problem for KVM switch cables.

Just as there are differences in rack depths, differences may exist in mounting holes. Although there are typically three holes per U, their exact makeup can be different. The end result is that some of the fancy mounting systems in use by the big server vendors may not fit as intended.

Perhaps the greatest difference in rack mountings lies in whether the unit already has predrilled and threaded holes, or, more likely, if it uses square holes and requires a set of rack screws, which includes rack clips. Rack clips contain a nut and affix to the inside of the square holes in the rack’s mounting areas.

Of course, the biggest problem is not necessarily getting things to fit. Sometimes it's keeping the rack from tipping over. Despite its rather steady appearance, it's quite possible to tip over a rack cabinet. It is for that reason that there are two essential steps to follow when mounting devices in a rack cabinet.

First, make the rack heaviest on the bottom. You should place heavy items in the bottom, and load the rack from the bottom first. This typically means loading UPSs into the bottom-most positions of a rack, if UPSs are used. It also means that servers should be installed in the bottom of the rack first.

Author’s Note
When planning a rack, don't forget to leave space for the keyboard, mouse, and video display. These are a few of the items that do not have to be loaded from the bottom up. Similarly, tape drives or tape libraries requiring access should be mounted in positions other than at the bottom of the rack. A more centrally located position will make swapping tapes easier.

Second, secure all devices in the rack. Once you place the servers on the rack rails and slide them into place, it's easy to forget to secure them into their final positions. However, if a rack begins to tip over, failing to secure the servers could accelerate the fall by shifting the rack’s center of gravity.

In addition, most rack cabinets include plates that allow them to be bolted to the floor. This is an excellent way to add stability, especially in situations where the rack is directly on a concrete floor.

With the high density that servers are reaching today, it's easy to see how they are increasing their power consumption. Each server requires a certain amount of power to operate, and the more servers in a rack, the more power the center will consume.

All of this power must somehow be distributed to the computers. A single 15-amp power cord, like the ones used in our individual PCs, are not able to support the power needs of the whole rack, so you must make higher-capacity connections.

Amps, volts, and plugs
Power is measured in amps (the amount of flow) and voltage (the amount of potential). While most racks only require the 120 volts derived from a standard outlet, they almost always require more than the 15-amp circuit typically found in homes and offices across the U.S. Because of the difference in amperage, different kinds of plugs and outlets are used.

The first step up from a standard outlet and plug arrangement is a variant called a T-Blade. Instead of having two vertical connectors over a circular connector, it has a horizontal connector and a vertical connector over a circular connector. The top two conductors form a T. These plugs are rated for 20 amps instead of a typical cord’s 15 amps. Because of the blade arrangement, it's possible to have an outlet that can accommodate either 15-amp or 20-amp plugs. The outlets can be easily differentiated from a standard 15-amp duplex outlet because the top-left opening will also be a T to accommodate either a vertical or horizontal connector.

However, the problem with this kind of a connection is that it's very easy to unplug with just a quick kick or pull of the cord. That's why locking connectors are used on almost every connection above 120 volts at 20 amps. The locking plug is a set of three blades arranged in a circle. Once the plug is inserted into the outlet, it is rotated clockwise to secure it into position.

Locking plugs come in many varieties, but the most common variety is an L type. The numbering scheme for these L type plugs and outlets is simple. A 120-volt connector is an L5. A hyphen, then the number of amps supported, follows the L5. An L5-20, therefore, is a locking 120-volt, 20-amp plug and outlet. Similarly, a 220-volt connection has an L6 designation; a 220-volt, 30-amp plug and outlet are called L6-30.

A UPS or a Rack Power Distribution Unit (RPDU) is commonly supplied by a 20-amp (L5-20) or 30-amp (L5-30) 120-volt locking connector. These connectors are typically sufficient to feed half of a rack or an entire rack.

Once the power is in the rack, it's necessary to distribute it. This is typically the job of the RPDU or, more simply stated, the Power Distribution Unit (PDU). Occasionally, racks are installed without PDUs. Sometimes the outlets from the back of the UPS (if installed in the rack) are used. Other times, power strips are dangled in the rack.

The PDU is designed to take in a large amount of power and distribute it to several outlets. This is the same function as a power strip; however, the rack PDU is optimized for its purpose in two key ways.

First, I said above that racks are 19-inches wide. This is, however, only the internal mounting dimension for computers. The rack itself is slightly wider. There is a small amount of space on both sides of the rack reserved for cable routing. PDUs are designed to fit in these spaces. A PDU can be mounted with ease directly to the outer rack space, allowing for a cleaner cable management solution.

Second, PDUs are configured with the same kind of outlet you will find in the back of your PC. This configuration allows for a denser placement of outlets. A PDU may allow you to place eight or more outlets in the same space as a six-outlet power strip.

In addition, PDUs can be configured for higher amperage output, while most power strips are not designed to support more than 15 or 20 amps of current. Some PDUs can accept and distribute 30 amps of power.

Power paths
Power within a rack will basically flow one of two ways, depending on whether the rack contains a UPS or the power conditioning and redundancy are handled outside of the rack.

If the rack contains its own UPS, the power will go directly to the UPS for conditioning and redundancy. It will flow from the UPS to power strips or PDUs for distribution to the devices in the rack. Occasionally, there will not be power strips or PDUs connected to the UPS; instead, the devices will be plugged into the UPS directly. The other available power path used in the absence of a UPS is power sent directly to PDUs or power strips.

It is not advisable to connect PDUs to the utility power coming into the rack and then plug the UPS into the PDU. The net effect of this is that the PDU creates outlets that are not power-protected. There is only one device in a rack that does not require power protection—the LCD monitor. Other than that, all of the devices should have protected power.

All of the power going into a rack generates a lot of heat. Computers and other forms of electronics are damaged by excessive heat. While the devices within a rack have their own fans for pulling air through the device, their effectiveness can be limited if the air inside the rack itself becomes too warm.

Servers and other computing devices rely on convective cooling. That is, they force air across the warm components in the device. The exchange heats the air and the device is cooled. Three things determine the effectiveness of this process. First of all, the amount of surface area of the device—which is fixed—can determine cooling. The second factor is the amount of air passing across the device’s surface, which can be impacted by obstructions or fan failures. Finally, the differential between the air temperature and the temperature of the device may affect the effectiveness of the cooling process. The greater the differential in air temperature to device temperature, the more effective the cooling.

You can only control two of the three factors that affect cooling: airflow and temperature differential. Either one can only be controlled indirectly.

To control the amount of airflow, you must ensure that there are no obstructions in front of or behind the fans that are designed to pull air through a server. Obstructions are often caused by cables located too close to the fan, making it harder for the fan to pull air through. By simply moving a cable a few inches, you can greatly impact the ability of the fans to keep your computer cool.

The other method of controlling cooling is to control the ambient temperature, which is the air surrounding the servers. We all know to keep our computer rooms cooler than we would normally keep our homes. Sixty-five degrees is a common temperature for a computer room.

However, in a rack enclosure environment, there is another ambient temperature, which is the temperature inside the rack. Because the servers are in the rack as well as in the room, it is important to consider the temperature inside the rack. In some cases, the temperature inside a rack can be substantially higher than the room temperature. In those cases, you must take steps to improve air circulation within the rack.

There are two things that can be done to improve airflow through a rack. The first applies only if you have a forced air, raised floor air conditioner: You simply place a vented floor tile directly under the rack to allow air to be forced up through the rack.

The other entails adding a set of fans, which are designed to pull air up through the rack, at the top of the rack. The added airflow should lower the temperature in the rack and improve the ability of the server fans to cool.

Ready to build
Racks are not magic, but they do require a bit more knowledge than just stacking servers on a desk. After considering mounting, power, and cooling, you should be ready to build your first rack without worrying about your safety or the reliability of your systems.

Editor's Picks