What do you need to know about a data center, whether using it for hosting or co-location? Above all, of course, you must know that your data will be safe and that support will be there when you need it. More specific questions should often be asked before making your choice. A number of important ones are listed below. (Additional good questions to ask if visiting the data center are, “Where is the restroom?” and “May I please have a promotional T-shirt to commemorate my visit?”)
For this piece, I looked at the perspectives of Barbara E. Hernandez of PC World and William Dougherty of Data Center Knowledge.
1.) How physically secure are you?
IT security can be explored from two primary angles – virtual and physical. With data centers, since they will be the actual location of your servers, start out by considering the physical component. Barbara advises that data centers be housed in “a standalone building with at least a 20-foot, fenced perimeter and a secure, cool core of servers.” She recommends that the location not be directly adjacent to your office. It should also contain standard measures such as surveillance cameras and security guards. There should be two points of entry. Access should be granted by photo ID.
Employees at the data center should all be trained for emergency situations (rather than being reliant on managers or specialists). Emergency protocol should include the usage of back-up generators and immediate, crucial data back-up to another location. There should also be servers available at a different data center that can be made available if disaster strikes (such as a hot-air balloon crash or an organized infiltration by thousands of Romanian cyber-sparrows). The data center should receive electrical power from two different sources so that uptime is maintained in the event of a blackout or brownout.
2.) What parts of the data center are concurrently maintained and fault tolerant?
As William argues, you want your data center to properly maintain its equipment, but you also need to know that while equipment maintenance is in process, fault tolerance is still in place. In other words, while that maintenance is underway, if the power goes out or hardware otherwise fails, are reasonable backup checks still in place?
In the event of a number of problems occurring at the same time, make sure you are protected, at least beyond a single redundancy (bare minimum, 2N or N+2). William writes, “Your critical IT infrastructure operates in a world where utility outages or equipment failures happen.” Make sure that you have multiple layers of protection.
3.) How are you optimizing energy efficiency?
As we all know, energy efficiency is not just important to environmental sustainability – it is also a key way to reduce the cost of operating any high-power system. The management of the data center, according to Barbara, should always be open to ideas from customers and any new innovations that can minimize power expenditures. Barbara lists eBay and Facebook as frontrunners in this effort, forming a relationship with the utility suppliers for their data centers to enhance the effectiveness of their cooling technologies (“Blow on them every once in a while,” offered one official). “Granted,” she writes, “both eBay and Facebook probably have more clout than a small business, but your data center should be listening to all of its customers.”
Having the most up-to-date hardware can cut down on power costs as well. Newer, more energy-efficient models can reduce cost as much as 40%.
4.) What are the data center’s average and maximum power densities?
In the early stages of the data center industry, facilities were not built with as high of power densities as they are now. These centers put more room between cabinets to allow for increases in density as needed. Data centers typically allow between 100 W and 175 W per square foot. Newly designed data centers tend to be built for a range of 225 W to 400 W per square foot.
That covers the square footage power capacities. The power density of an individual cabinet must be examined as well. A decade ago, as William points out, cabinets could have a 2 kW limit. Now, you want your rack to allow for 8 kW to 10 kW. Furthermore, this increase in density maximum is still on the rise. William advises, “Expect your required power density to climb and make sure your data center has the infrastructure to grow with you.”
5.) How does your cooling system operate?
The core temperature of any room in which servers are placed should be between 68 and 75°F. Typically, about 50% of the overhead of running a data center is the cost of cooling it, per Barbara’s assessment. Even though it is so outrageously expensive to cool, many data centers overdo it, pushing the temperature below the minimum threshold of 68° (with the added bonus that the room can then also be used to refrigerate their groceries).
Many data centers now operate the climate control of their facilities through a control panel, with monitors and gauges throughout the buildings to ensure temperature is ideal and power bills are minimized.
6.) How are you protected against the common regional natural disasters?
Wherever your data center is located, it can be a victim of a natural disaster – including hurricanes, snowstorms, tornadoes, earthquakes, and wildfires. William lists one possible devastating scenario: “the data center survives a massive earthquake, but utility power is still out with no estimated repair timeline.” You want to know, under such circumstances, the length of time that the facility can continue to supply power via its generators with the fuel it has on-site. Furthermore, does the data center have a number of different suppliers available to bring in extra fuel during an emergency?
Make sure that you fully comprehend the most likely types of emergencies that might befall the facility. Based on that information, develop emergency plans to mitigate risk.
Barbara’s commentary on this issue is also worthy of mention. She cites the 2007 power outage that brought down 365 Main’s facility in San Francisco. 40% of the data center’s customers experienced downtime that lasted for about 45 minutes. The problem in that case was that not all of the generators kicked on as intended. Eight were required for backup purposes, but only seven started due to a malfunctioning electronic controller.
What 365 Main’s clients learned from this, says Barbara, was “to find out how a data center plans to notify customers in an immediate emergency, keeping them apprised of latest developments and the status of their company data or services.”
7.) What skills and training do the remote hands and eyes team have?
Servicing of your hardware will need to occur at regular intervals. There are two ways to perform this maintenance, says William: visiting the data center yourself or taking advantage of the data center’s remote hands and eyes technicians. You probably do not want a security guard performing this task, as sometimes occurs. The remote hands and eyes team should consist of individuals with IT credentials. You want to know what the requirements are for attaining that role. Speak with the person in charge of daytime and nighttime support. If a remote team is credible, your physical closeness to the data center’s location becomes less of a concern (if not, always be within a 400-meter radius of the facility, and wear your binoculars).
8.) What is the Data Center’s virtual to physical machine ratio?
How virtual is the facility? Virtualization is a good way to reduce your expenses. Barbara points out that a ratio of four-to-one allows server expenses to break even. However, the majority of servers can support as many as a dozen virtual private servers (VPSs). Maximizing virtual possibilities means higher efficiency regarding hardware, power, and rack space – so its cost-effectiveness is manifold.
9.) How frequently are generators load tested?
Load testing of generators is expensive – both because the equipment for testing is costly and the high amount of fuel used, says William. Data centers sometimes use a power outage itself – which is obviously not a test situation – to check if there generators can handle loads or not (which is also not a good time to load test a clown car). This routine maintenance is essential so that generator issues are discovered prior to emergency situations. You want your facility to give each generator an extended load test every quarter at the minimum.
10.) How is the data center certified, and is it audited each year?
Certifications are a simple, standardized way for a data center to provide you with credentials. As William states, “This information is invaluable because it represents an independent analysis of the facility’s quality, reliability and security.” If your website takes payments, for instance, you want your data center to be PCI-DSS compliant. Financial firms require SSAE 16. If your business is green-friendly, the LEED Gold and Energy Star certifications are crucial. Verify your data center is legitimate by asking for documentation of any certifications they claim to have.
Comparing datacenter options is a rigorous process. Knowing some crucial questions to ask can keep your data in safe hands even if it is not at your own immediate fingertips.
When reviewing whether a data center is the right choice for you, first, look at its physical security components. Then consider the relationship between concurrent maintenance and fault tolerance, along with its energy efficiency. Ask about its power densities, cooling systems, and contingency plan for natural disasters. Know the skills of the remote hands and eyes team, the virtual to physical ratio, and the load testing schedule for the generators. Finally, check on certification and auditing. Once you have all this information, you will have performed due diligence and are prepared to make a wise decision for your IT infrastructure.
by Kent Roberts and Richard Norwood