The Essentials of Data Center Design
Table of contents:
Your Business’s Central Nervous System
The Colo Option
Planning your Data Center Layout: Save Some Space
Know Your Power, Keep Your Cool
Backup Your Backups
Don’t Get Too Wired
Build Expertise Into Your Data Center Budget
Securing Your Investment
Get Virtual With It: Remotely Managing Your Data Center
Keeping Things Running: Ongoing Expenses
Putting Plans Into Action
A modern business has to grow and adapt like a living organism. Sales, customer service, advertising, invoicing, and product delivery are all part of one organic process catering to today’s consumer in windows that sometimes only last a few seconds.
If businesses are like living, moving things, the modern data center is the central nervous system that makes everything work like it’s supposed to.
Consider a common transaction: an ad on Facebook steers a user to the Acme Widget Company’s website, where the user shops the online store, selects a gizmo, calls an account representative to ask a question or two, then pays online with a card. The product then ships.
It seems like a seamless, fully automated system, but all kinds of things are happening in Acme Widget’s data center. Web servers are integrating application servers and payment processing software, while your Voice over IP (VoIP) server and messaging system handle phone traffic The whole process operates under the hyper-suspicious eye of firewall/IPS devices, which are in turn watched by the (hopefully) even more paranoid eyes of Acme’s IT infrastructure team. That transaction that seems so seamless isn’t, and any of a myriad maladies impacting your data center will prove it in a hurry.
By considering the following areas critical to planning a modern data center, you can preempt a world of problems down the road.
The complexity of data center operations makes them an increasingly attractive package to outsource, and massive data storage facilities called “colocations,” or colos, are a growing business.
Colocation spaces help by:
- Storing and managing your hardware
- Ensuring data connectivity
- Providing a secure and climate controlled environment, and
- Complying with all state and federal regulations.
But colos charge premium fees for everything from cabinet space rental to “remote hands” services, which involves tasking an onsite data center employee with simple chores like server reboots.
Colos are designed as data fortresses. Their redundant power and data, and impeccable security make them an attractive option when considering your new data center. But the size, scope, and demands of your business must be weighed against the costs of colocation.
Fires. Floods. Blackouts. These are the environmental assassins plotting against your data, so you want to place your data center in the most climatically boring area to which you have access. Acres of storage space are worthless if your space can’t be kept cool, clean, dry, and lit with data and power.
The increasing power of multiprocessor, enterprise class servers, coupled with the rise of virtualization technology, have made it possible to amass more computing power into less square feet than thought possible even five years ago. But when renting colo space or planning for natural growth cycles in-house, planners frequently underestimate how much space is needed to run an efficient data center.
Two common mistakes leave data centers short on space:
- Over-estimating racking equipment. Measured in units, or “U’s,” a six-foot cabinet may boast 42Us, but that doesn’t mean you can cram 42 devices into it. Variables like the devices’ heat signatures, available cooling, and required ease of access can gobble up rack space. To keep things orderly, plan on using about half the U-rating for actual equipment. So if you have 24U of devices, don’t assume a 30U rack will do the trick. Plan generously for cabinet space.
- Over-looking the need for human workspace. Poorly designed centers cram too much stuff into too little space. This makes basic maintenance about as easy as trying to play baseball on a submarine. Don’t just plan for the racked equipment’s space, but also for a practically-sized workbench where your pros can troubleshoot, monitor, and upgrade systems.
A good baseline for space planning is the 10:4 ratio. For every ten square feet of gear, add four feet for people to move and work around your critical equipment. Square footage costs money, especially in a colo, but having too little impacts operations costs more.
We already know that your data center will be somewhere with highly reliable (and redundant) data and power feeds. But within your data room itself, HVAC and power consumption can interfere with each other. Make certain the power circuits supplying your center are massively overbuilt, allowing for powerful data systems that throw off enough heat to warm a room and the HVAC machinery required to counter it.
But available voltage is only part of your planning. With the rapid proliferation of individual devices likely to pop up in your data center, there will be additional considerations like:
- Firewall hardware
- Storage (SAN/NAS)
- Backup arrays
Finally, it’s easy to underestimate something as simple as the number of outlets your room will need. In some cases, surge protector strips can extend a single outlet to many, but you don’t want a hydra’s head of extensions snaking around your center. It’s worth doing a count of the total number outlets you have in play now, then planning to double them for a center prepared to grow as you do.
Leave physical and fiscal room in your data center for uninterruptible power supplies (UPS). Industry leaders like APC provide a range of options for how long you can keep things running on battery power if a major power outage hits your site.
Lower-cost UPS options generally offer 30 minutes or less of emergency power—enough time for your systems to undergo a “graceful shutdown” rather than the hard crash of a power cut. By scaling up the size, power, and price of your UPS design you can increase the window in which your systems can go without AC power.
Just don’t overdo it with UPS. Remember these units are bulky in terms of both space and power consumption. Also consider that there’s little point in investing massively in UPS that will keep you running for days if a catastrophe has downed a major power grid and all the data connections that rely on it. A few hours of UPS battery backup is usually a reasonable insurance policy.
Electrical wires are the minority in a data center. Network cables and patches can seem to grow like hair, and many’s the new network pro who has walked into an indecipherable tangle of cables.
When planning your cabling, constantly strive for simplicity. Outgoing network cables should be clearly labeled, and ideally color coded by the areas they serve. For example, red CAT-6 cables feed the first floor, blue the second, etc. Uplinks that serve separate switch closets should have also their own color and be easy to identify.
All of these wires add up quickly and occupy more space in your data center than you might expect. It’s important to build in tidy, accessible wire management paths, either under a specially designed elevated floor or along specially designed tracks. While a true wireless high speed future may lay somewhere out on our horizon, for now, respect the Cat-6 cable. Failing to do so will create a room full of stress-inducing spaghetti.
A modern data center is no longer an unused closet in a corner. Its complexity should mirror the value you place on your business operations and data, so be prepared for some sticker shock when planning. There are several niche experts you will want to involve:
- Architects/Interior Designers—The architectural considerations of a data center are daunting. More than just another space in a building, many centers are designed on the elevated floor option, which builds in a building code-compliant space below the room walkways to route and organize your network cables, requiring specialized architectural design.
- Specialized HVAC Pros—A data center must be designed with the special power and HVAC systems outlined above, requiring licensed electricians as well as heating and cooling professionals. A quick web search in your area will turn up companies that specialize in data center construction.
- Low-Voltage Experts—And don’t forget the low-voltage pros. Electricians and data wiring experts aren’t always the same group. Seek out low-voltage design companies with experience building clean, impeccably organized network and punch-panel arrays. These will always include clear maps and manuals outlining what network drops go where. Construct a low-volt wiring scheme an amateur could read.
- Data Center Fire Suppression—Any normal and code-compliant business space includes fire suppression, but a sprinkler system won’t do it in your data center. In addition to the downside of soaking your hard drives and electronics in water, the protective sheathing around data wiring can emit dangerous chemicals when burning. To counter this modern data centers employ an argon gas fire suppression system which will put out a fire in a hurry, but cost thousands of dollars to install and maintain.
Just as you don’t want to skimp on physical space for your data center, be generous in its construction and maintenance budget. Today’s center depends on experts across many fields, so plan for their cost accordingly.
It goes without saying that after all the effort you’ve invested in creating a data fortress, you want to guard the gates against intruders. A random wanderer in your data center could pull the wrong power cord and wreak havoc. A malicious intruder could engage in espionage or worse. Here are some tips for keeping things locked down.
- Create an access list. Before the room is operational it should be off limits to everyone not on a master access list. This list usually includes only IT directors and select technicians. Maintenance and janitorial master keys that work in other places in the building can’t work for entry to your center. Unless specifically mandated, the CEO should not have access, as generally there is nothing in there that he or she should have to manage.
- Use electronic locking systems. Good old-fashioned deadbolts are a great backup, but a digital locking system that requires keycard access is better. A mixture of hardware and software, electronic locking systems control your master access list and tell you who is moving around in your center, what time she’s doing it, or where he’s been. These systems keep a master log that can be accessed against events to determine if human error, or something darker, is impacting your data center.
- Use camera surveillance. Today, camera surveillance and recording technology is inexpensive. Building in some strategically placed IP cameras will offer the option of taking a real-time peak of the data center from your smartphone. Look for camera package with an authenticated time stamp and a resolution rate high enough to be considered as legal evidence in your data center’s state(s). Most of your tech headaches will almost certainly stem from other issues, but if someone is up to no good, make sure the authorities can see him in the act.
If you really want to get state of the art, build in a biometric security system. This takes the electronic monitoring of the master access list to a new level, eliminating all doubt about who can and cannot enter your data center by basing entry on fingerprint or retina identification. It adds an advantage: the harried systems engineer responding groggily to a three am emergency won’t find himself at the data center and his access card at home.
With your security tightly buttoned, it’s time to work on saving you trips to the data center by planning for remote management.
If you’re not already using a hypervisor you soon will be.
“Hypervisor” is the collective term describing virtualized network management suites, the most popular of which are offered by VMWare and Microsoft Hyper-V. Hypervisors control your virtual network ecosystem and come with management tools that let you log into your data center from anywhere and manage servers, virtualized switching, and even the individual desktop experience of remote users.
Collaborative efforts like the Xen Project aim to offer free, open-source hypervisors. As with all things open sourced, the obvious savings must be measured against variables like the availability of emergency help, the frequency of fixes and security patches, and vulnerability to outside attack.
We are entering the next evolution in the virtual age. The challenge for IT leaders will be leveraging the vast power of virtualized network and computing environments against the ability to control and maintain them.
Once you’ve achieved your perfect data center, get ready to change it.
Things move quickly in the digital world and planned equipment life cycles of three to four years might not be fast enough. Given the massive workloads that modern data centers are expected to carry, it’s important to make sure you provide enough hardware horsepower to meet the challenges.
This starts with bandwidth. Once upon a time a 1.5Mbs-dedicated connection was enough to run a business. Today it may not be enough to run your VoIP phones. As innovations in virtualization reduce user-side hardware profiles, the throughput needed to seamlessly serve applications from your data center will grow exponentially. Consider your current data usage and growth estimates, and plan on growing your bandwidth accordingly.
Remember when the 1.44MB floppy was the standard unit of storage? Today that wouldn’t hold a single selfie snapshot. Modern data consumers do just that—consume. Pictures, video, music, and more fill up space almost as fast as it’s provided. In addition to sensible storage limits for your users, be assured that your storage arrays—and the backup systems duplicating them—will need frequent upgrades. Disk space is cheap and getting cheaper, which is good, because you are always going to need more of it.
With all this talk of construction, wires and hardware, don’t forget the humans. Data centers don’t maintain, upgrade, and repair themselves. As your operation grows the need for talented people to shepherd your flocks of machinery must grow with it. Staff with the capability to deploy instant crisis response 24-hours a day. Hosting your data center at a colo comes with the option to call upon its 24-hour staff, but remember that “remote hands” services are much pricier than in-house staff.
Any organization that wants to be competitive in the digital world relies almost entirely on the strength of its data center. The above are some of the key considerations you’ll face when plotting changes, but there’s always room to invest more research and planning into your data center.
With proper design, planning, expertise, budgeting, and oversight, you can build a strong, agile, digital central nervous system that empowers your organization to respond instantly to challenges and opportunities.