Whether you’re only looking for half a cabinet or carefully laying down a cross-continental deployment strategy, choosing a data center is never a trivial task.
In fact, with most contracts signed for a pre-determined term, this is likely to be one of those make or break decisions that can either boost or cripple your business. Once the hardware is in place, relocating it is both complex and time consuming. As a result, implications of a wrong choice could stretch far further than the immediate loss on a bad investment.
To help you avoid these bad decisions, we present a list of tips for choosing a data center, drawing on lessons learned through years of worldwide server deployments.
1. Take the Tour
This may sound obvious, but you’d be surprised by how often racks are purchased solely based on the vendor-provider information or audits.
So make the effort, buy that plane ticket, book that hotel and take some time off to get your boots on the ground. When you do, even when accompanied by an auditor, come prepared with your own checklist and don’t hesitate to ask hard questions.
2. The Staff
Whatever can go wrong, will go wrong. When it does, the local IT staff will be your arms, legs, eyes and ears, as you work to resolve the issue. This is why it’s so important to know what the response time SLA is and to specifically inquire about the number of technicians who are on call 24/7.
3. The Gear
We bring our own hardware but there many reasons why you would prefer to use data center gear. If you do, make sure to go through all the specs to understand exactly what computing and networking gear is available.
Finally, it’s always a good policy to ask about equipment refresh rates, especially if you plan to stay with your provider for a while and don’t want to get stuck with an outdated hardware.
4. High Connectivity
It always makes sense to know which carriers have presence in the data center. If, like us, you are in the business of optimizing content delivery, your focus should be Tier-1 providers with our own short-list including: Level3, Telia, NTT and Cogent. For redundancy, we always suggest connecting to at least two different carriers; more if you expect bursts, like the ones you get during DDoS attacks.
Another thing to consider are the public peering options available in that facility. With the right peering agreements you can improve your performance and significantly cut down on transit costs.
5. Security Standards
Data center security is a clear priority. It starts with physical security measures, which should be a part of your checklist, as you take a tour around the facility.
Things to pay attention to include:
- Buffer zones around the facility.
- Gates and barriers that limit access to the perimeter.
- CCTV cameras in access point and strategic locations.
- Mantraps and fingerprint scanners.
- Redundant utilities (water, air, communication, and more).
- Sturdy private cages.
Physical security measures aside, you also need to be aware of the security standards that apply to your industry and inquire about those, to ensure compliance. The most common example here is the PCI DSS (Payment Card Industry Data Security Standard), but there are others you may need to be aware of as well.
6. Bandwidth and Bursts
Check your SLA for bandwidth limits and burst costs. Knowing these will help you avoid hidden costs and possible downtimes as a result of traffic spikes. More importantly, planning for these in advance could help you to better manage your traffic and bandwidth costs, by balancing loads between multiple facilities.
7. Power Management
On average, power and cooling costs account for nearly a quarter of all data center operational costs. The reason why you, as a client, should care is because those extra costs will filter down and impact your own prices. This is why it’s important to ask about energy efficiency policies and improvements which may or may not occur down the road.
8. Energy Storage
Blackouts will occur. When they do, make sure that the facility has an effective backup plan, which goes beyond just having connection to multiple power grids, as in the event of power shortage all of these are likely to be affected.
This is where energy storage comes into play, with options including: batteries, flywheels or hybrids. You should know which ones of these are available in your facility and for how long they are expected to keep it running, when the main grid goes down.
9. Raised Floors vs. Slab Floors?
This argument’s been raging for a while and we have no plans to settle it here (nor will we reveal our own preference). What you need to know here is that both options have their faults and their merits. Also, you should be aware that slab floors require overhead cooling and cable hookup, while with raised floors cable and cool air will be passed from underneath the cabinet.
Raised floors are likely to fair better in case of floods, while slabs offer more stability against earthquakes, so—statistically speaking—some regions could be a safer choice for one than the other.
10. A Name You Can Trust
Your SLA is only as good as the organization that stands behind it. To be on the safe side of things, we always prefer to work with an established colocation provider. When this isn’t an option, commit time to in-depth research and don’t be afraid to ask the provider for references.
Better yet, go through the customer list and then use LinkedIn to find a CTO of similar-sized company, who already hosts in that facility.
11. Out-of-Band Network Management
High availability comes at a cost. Out-of-band (OOB) management is a dedicated channel that lets you remotely control your equipment when disaster strikes and all your regular in-band options become unavailable.
Is it worth the extra money? Only you can answer that question but we believe it does, especially for business-critical infrastructures. Also if, like us, you offer an enterprise-grade SLA, OMB should be a no-brainer.
12. Location, Location, Location
If, like us, you are in the content delivery business, the cliché first rule of real estate applies and the process of choosing a data center always starts with finding the optimal deployment geolocation.
If you have multiple options, often across several continents, the first step is to observe traffic patterns across your own network to identify high-demand regions, that could best benefit from additional coverage.
At Incapsula, we use several monitoring tools that help us do that. One of the most interesting tools is a home-brewed solution that sends DNS requests to publicly accessible DNS servers, each of which pings back with a response to the nearest Incapsula PoP.
The image below shows how, by mapping these responses, we were able to identify a high-demand spots in Central Europe (Poland, Belarus, Ukraine and Western Russia), which were served by a somewhat remote Stockholm facility—indicated in pink on the map.
This led us to investigate regional data centers options, finally resulting in a new PoP in Warsaw TelecityGroup’s facility, chosen based on a combination of the aforementioned criteria.
Consequently, in the second image you can see how the new PoP helped us optimize our coverage in the region, while taking the traffic load off the Stockholm PoP.
There’s an interesting anecdote about this monitoring tool: All the DNS servers it pings are machines previously used by hackers to launch DNS amplification DDoS attacks against our clients.
Our IP reputation system recorded the addresses of those public resolvers and started putting them to a good use, giving a 21st century spin to “turning swords into ploughshares”.
Want to learn more about our data center deployment strategy? Contact us or leave a comment below with questions or share your own tips for optimal data center deployment.