Sustainable Information Technology Series – Server Room & Data Center Setup

The proper set up of a server room is certainly not on the top of most small businesses lists of mission-critical issues, although perhaps it should be. Over the years, we have seen many server rooms and data centers, some good and some are not so good. (The worst we’ve seen was sharing space with a men’s room!)

What most surprised us was seeing how some medium and large businesses operate their server rooms and data centers. Some may appear to be well designed and maintained, but this façade is quickly revealed when the power or HVAC fails. Even the best-designed data centers that follow all best practices have failed, despite all their built-in protections. And some have failed repeatedly over the course of a few months.

Large Data Centers

You have probably seen depictions of data centers either in person, if you work in high-tech, or if not, in the movies. The archetypal data center has a raised white floor, is temperature controlled with a sophisticated fire and smoke detection system, moisture sensors, controlled power with an uninterruptible power supply, and huge bandwidth to the Internet. Of course, data centers also have very heavy-duty security control systems with electronic key locks, 24x7 security guards, and surveillance cameras. This is because the data center security mantra is well understood by IT professionals: physical access = full access and full access = zero security.

Raised Floor

Why is a raised white floor part of the ideal data center? A raised floor performs three functions:

It prevents flood: Whether caused from external, environmental factors or something as simple as a broken pipe, a raised floor is key to reducing equipment damage. In a minor flood, where water doesn’t reach the equipment on the elevated floor, an electronic short will trip the circuit breakers, preventing a major incident. Of course, the floor will not prevent electrical or equipment failure if the water gets too high. If your data center gets hit by a tidal wave, you’re out of luck.

However, equipment sitting on the floor is still in danger. It can short out before the circuit breakers have a chance to trip by something as small as a quarter inch of water from something as simple as an air conditioner leak.

It maintains a clean, organized environment: It is much easier to keep the center organized and clean with a raised floor. Think of all the connections needed to and from the equipment and electronics in the room: power cords, telephone lines, and data equipment wires and cables are in abundance.

With a raised floor, all these can be run under the floor and easily accessed by removing the floor tiles. Cords run between wall plugs to Uninterruptable Power Supply (UPS). Power cords run from the UPS to the equipment; network cables run from all the switches, routers, and telephone equipment, in addition to keyboard, mouse and video cables.

Imagine if you didn’t have a raised floor: wires are exposed and anyone can trip over any cable coming out of the server. Imagine this happens with your server that holds, for example, all of your accounting information, causing a panicked recovery and ruining your day.

It controls the flow of air: Air-conditioning can be run under the equipment as well as over it, providing more options for airflow management. Holes drilled in the tiles allow cool air to come up wherever you want, just as it does from the vents in the ceiling. Also, by putting the equipment in a housing that has fans on the top, cool air gets pulled from the floor up to the ceiling vents.

Air flow and A/C

Why is A/C so important? Heat is the enemy of electronic equipment: the hotter it gets, the shorter its life. A server kept at 86 degrees may last only half or a third as long as one kept at 75 degrees. Operated in 90+ degree rooms, servers can fail in days or even hours.

While the server specifications indicate a very broad range of operating temperatures, operating at the extremes is not recommended, and what is not indicated is how much a server’s life is shortened as it approaches either end.

Consistency of temperature is important. By increasing and decreasing the temperature, it is possible to pop the chips off the circuit boards, crack the soldered connections and generally wreak havoc with components. Think of thin bits of metal heated then cooled over and over and how likely they are to warp or break from the stress.

Therefore, data centers are kept in an air-conditioned environment, preferably below 65 degrees. If the room is not full of equipment, the center can become quite a bit colder, since the A/C is normally planned for full capacity. In some cases, the thermometer can only be set so low without affecting the airflow in the room. You can have too much A/C: it can freeze the ducts and turn off before humidity is removed from the air, so condensation can form on components for a self-created flood or at minimum, moisture damage.

A/C output is measured in Tons, and heat output of equipment is measured in British Thermal Units (BTU) per hour. Something to keep in mind: 1 Watt of power generates 3.4 BTU/hour; 1 Ton of A/C handles 13,000 BTU/hour, about 3600 Watts of equipment.

As well, you’ll notice that there are frequently no windows in data centers since sunlight can play havoc with your A/C calculations.

Airflow matters as much as A/C cooling power. Data equipment does not do well in uneven temperatures, and good airflow prevents hot and cold spots. Consistent airflow also keeps the heat that the equipment generates moving and dissipating, keeping the room temperature even and consistent.

Power

Large data centers primarily house data racks. Data racks are measured in Rack Units (RU). A typical full data rack or enclosure has 42 RU, each about 1¾ inch, making the rack a little over 6 feet high. Rack-mounted equipment can improve organization by keeping everything bolted down, making the equipment more stable. In some cases, the racks have slide-out rails, making the equipment easier to service.

Very large data centers have conditioned power to the whole floor or the whole building, with UPS and standby generators. Downtime is costly, and sudden downtime is deadly.

Power ideally comes in from two totally different power grids, either of which is able to handle the needs of the entire data center. The amperage can be in the normal range of 15 or 20 amps, or if the equipment is especially powerful, 30 to 100 amps with circuits at 110 or 220 volts. Circuits can generally be stepped down in amperage or voltage, but stepping up can be much harder, since at some point you may run out of power.

Bandwidth

To get high-speed access, many computers require large throughput connections to the Internet, but by dividing the cost as well as the bandwidth, these connections can be justified. For example, a DS-3 line at about 45 Mb/s can be shared in a data center by 100 computers, allowing each to have a minimum of.45 Mb/s. The fewer computers, the more Mb/s each can have. T-1 to OC-255 can be contracted for guaranteed uptime and bandwidth, and for most data centers this is worthwhile.

Server Rooms for Small to Medium-sized Businesses

Many of our clients are small to medium-sized businesses that need the features and benefits of a full-blown data center, but think that they don’t have the space or the means. They’re wrong. Many aspects of a large data center can be scaled down in cost-effective ways. For example, a small room can serve as a small data center, sometimes called a server or computer room.

Below are some critical elements:

Security: Security practices include locks on all doors and limited access to only key personnel. Security monitoring can be inexpensively provided by a camera with an IP address as a network camera, and as well, other sensors are surprising reasonable.

Power Requirements: If possible, get the office computers wired on separate circuits. At minimum, don’t put your computer(s) on highly variable power, for example, on the same circuit as a refrigerator or elevator. Refrigerator condensers and elevators can draw huge loads intermittently.

Get one or more Uninterruptible Power Supplies (UPS). We recommend getting a model that runs on a battery with a backup wall plug that keeps the battery charged. These are the most reliable at providing predictable power, although not always with a clean sine wave output. UPS are typically rated at 60% of their sustainable capacity, so a 1000 Volt Amp (VA or 1kVA) unit will handle 600 watts of equipment.

If you need to bring an electrician out to help with wiring and you anticipate company growth of your server room or data center, you should consider bring in multiple amperages for future needs. 20 amp outlets are keyed for both 15 and 20 plugs, which will be adequate for most computers. Some high-end servers, networking equipment and SAN or iSCSI based storage arrays require 30 amps. Normally, unless you’re using Enterprise level equipment, more amperage is not needed.

Communication links: All modern server rooms require good communication links, especially data communication links. Ideally, the building where the server room is housed would accommodate more than one type of data communication link, such as DSL, T1, Wireless, Ethernet or Fiber to Internet.

Temperature controls: Adequate A/C and air circulation needs to be provided to the room. Because this is so key, it may be worthwhile to bring in an A/C consultant familiar with the needs of small server rooms.

Some factors to consider are:

	Add 4000 BTUs for each room below a ceiling or roof which is not insulated (not recommended).
	Add 1500 BTUs for each window which receives significant daily sunshine.
	Add 1500 BTUs if room is above a boiler room or kitchen (if it is in use).
	Add 600 BTUs for each person in the room. You can subtract 1000 BTUs if you are on the shaded side of a building.

	Reduce short and long-term capital spending on additional hardware and software which would be required to provide the same level of SSOC services.
	Reduce time spent on overall IT maintenance issues. It has been estimated that maintenance issues can consume 50% of total IT time.
	Reduce errors made in performing IT maintenance by taking advantage of the TNS procedural approach and expertise.
	Reduce staff time spent after hours and on weekends by taking advantage of the SSOC 24x7x365 operation.
	Ability to have 24x7x365 real-time monitoring and notification of:

	Network issues
	System issues
	Application issues
	Security issues
	Environment issues

	Ability to monitor your critical equipment and server room temperature (-40°F to 212°F), humidity (5% to 95%) & wetness (1% to 100%).
	Ability to have 24x7x365 real-time corrective action taken by our SSOC engineers, remotely or on-site.
	Ability to patch your critical OS and applications after hours and on weekends by our SSOC engineers remotely or on-site.
	Ability to restart your mission-critical applications remotely or on-site.
	Ability to track the possible network intruder by our SSOC engineers remotely or on-site.
	Ability to provide up-to-date network documentation (Green Book) for your review.

	Security
	Power requirements
	Communication links
	Temperature controls
	Physical layout
	Monitoring & management systems