About This Case

Closed

16 Dec 2009, 11:59PM PT

Bonus Detail

  • Each Selected Insight
    Earns a $100 Bonus

Posted

4 Dec 2009, 6:19PM PT

Industries

  • Enterprise Software & Services
  • Government / Politics / Global Issues
  • Hardware
  • IT / IT Security
  • Real Estate
  • Telecom / Broadband / Wireless

Picking The Right Spot For A Data Center

 

Closed: 16 Dec 2009, 11:59PM PT

Earn up to $100 for Insights on this case.

If you haven't noticed, we're starting an ongoing series of cases here to develop interesting, engaging and useful discussions for our new sub-site, IT Innovations.  We're looking for insights that might help IT managers stay informed and keep their operations competitive.  

So for this case's topic, we'd like to delve into the subject of picking a data center location. We're not too interested in lengthy narratives about horror stories in selecting a server farm back in 1998.  But that's not to say we don't want to hear your personal experiences.  We're just aiming more for concise, non-generic advice that might help your fellow IT gurus (or gurus in training).  A few hundred words should suffice.

If you're not exactly a writer, you could submit something like a mashed-up map for where data center locations are optimally located -- just as long as your contribution is relevant and useful.  A picture can be worth a thousand words -- and is sometimes more illustrative of a point.  

And this topic isn't restricted only to geographic factors.  There are likely a multitude of reasons for choosing a data center, and we'd like to hear what you think are some of the most important.  

 

7 Insights

 



Several factors apply to a good data center location.

Natural Disasters - The location should be resistant to the common natural disasters.  Avoid the coasts (storms, tidal waves), avoid floodplains, avoid areas of high tectonic activity (earthquakes and volcanoes) and if building in areas impacted by large thunderstorms and tornadoes, build underground.

Ease of Access - This must be balanced with the above, but if you are in a highly secure location, it often means that you are harder to get to.  This makes it more difficult to get server equipment and emergency personnel to the facility, which can impact how much the location is used.  Granted, if you are not reselling space, this is less of a concern, but it remains a factor.

Local Providers - A data center is likely to need power and connectivity.  The local utilities that can provide this must be willing to work with you.  Some areas have very difficult laws in place relating to backhaul, and that alone may raise operating costs enough to impact profitability.

Local Politics - If the data center is likely to be large and especially if it could cause job creation, it's worth seeing if a local government will offer tax breaks for selecting a location.  These sorts of games can slow the process significantly, but it can also save millions of dollars.

Local Keystone Client - Odds are that there is a local large employer that is in need of new space.  There may be cost savings involved in sharing the development costs and building a room in the new data center that is dedicated to the keystone client.  If possible, find a way to charge for usage at such a rate that the one client pays the operating expenses, so if pickup for the rest of the datacenter is slow, it doesn't impact the sustainability of the business.

Local Management - The good locations tend not to the same "hip" locations that attract the best operations talent.  Be prepared to pay above market rates for staffing in order to get the level of professionalism required.  Remember that you'll need, at minimum, one person doing the business-side, one person available for facilities issues and one person per shift to deal with network, power, storage and server issues.

I am interested in the space where science and art combine. I thrive in the intersection of business, security, and information technology, and enjoy helping clients achieve balance by reducing risk, accepting the risk or altering processes.
icon
Don Bradley
Mon Dec 7 1:59am
>...it's worth seeing if a local government
>will offer tax breaks for selecting a location.
Its probably cheaper in the long run to select a locality wherein the local government not only will not grant tax breaks but will not even consider granting tax breaks.

Once municipalities start trying to attract industry by granting tax breaks, they over burden their infrastructure. They give away the tax revenue needed for schools, roads, parks, etc. and the quality of life deteriorates. Then you don't have a pool of talented well-educated programmers available to you because few professionals want to live in an area where roads are clogged, hospitals are not being built, parks are not being maintained, etc.

Its similar to getting an airline ticket for too low a price. If the airline is not charging you a high enough price, then they can't fly you safely. If a municipality is not getting sufficient tax revenue they can't provide an attractive place for technicians to live and raise a family amidst rising real estate values. If industry does not shoulder the tax burden, over burdened Single Family Homeowners will have to do it. And there goes your talented, well-educated work force!

The Choice of a Data Center location is largely influenced by emotional need for security, Business Sense and Technical Prudence.

As one of the largest data center provider of India, our team is constantly juggling between these factors.

It is important to understand the emotional needs of user community. Even in this age of borderless commerce and cloud based virtual infrastructure; there is still a premium for secure and accessible data center facilities. Users continue to feel safer if their servers are close to the work location. They are willing to pay a premium for accessible and secure locations. In addition to emotional needs, there may be regulatory and political factors for locality of the data. The recent drive by government of India to fund State Data Centers(SDC) in many of the Indian states under the National e-Governance Program is a visible example to fulfill these needs.

In my Experience, It was easy to convince the datacenter users/customers to move-in the computing assets after a brief tour of the data center. I am assuming that this experience gives them a sense of security and accessibility to trust their assets to the datacenter providers.

It makes perfect business sense to evaluate initial investments, Input costs and consumption patterns before deciding the location of the data center.

The real estate cost, construction viability, un-interrupted availability and consistent price for electrical power play a vital role in choosing the location of the data center. Some Projects are being conceived in locations close to power sources such as hydro-electric, thermal, and other power plants to reduce the power distribution losses leading to lower power tariffs and increase in viability/profitability.

 An important factor for Internet Data Centers is the proximity to Internet Exchanges (IX). Multiple access paths, reduced latency and high bandwidth internet connections allow building of reliable high performance VPN based overly networks for the datacenter users.

The climate of the location plays a crucial role in determining the HVAC needs and drives the associated costs, Dry and moderately cooler locations have reduced HVAC costs.

In addition to HVAC, virtualized multi-tenanted design of compute, storage and network infrastructure determine the level of commitment to meet the Green IT objectives.

Green IT may not only be statement of corporate social responsibility but also allows to leverage local government incentives tied to meeting the Green/Environmental objectives.       

Picking The Right Spot for a Data Center helps in successfully building a Secure, Powerful, Well Connected, Cool and Green data center of the future.

http://www.sifycorp.com/scripts/datacenterservices.asp (our team) 

 

 

 

 

Where I might build a datacenter is a pretty vague question depending on a ton of factors.  A much easier question to answer is "where would I not want to build a datacenter?" 

1. I would not build a datacenter anywhere that I couldn't get to and physically set foot in the datacenter within a few hours.  I don't care how cheap it is, how do I know if I'm getting a good value if I can't drop in unexpectedly and see if it's being run as planned? 

2. I would not build a datacenter in a zone known to experience frequent or severe natural disasters.  Again, what's the point of saving a few bucks up front if a disaster (tornado, earthquake, hurricane, etc.) can wipe me out? Furthermore, how stupid am I if I put my datacenter in a place known to experience those things?  The Earth is quite large, I can just put it somewhere else.

3. I would not build a datacenter anywhere that had a single point of failure on a necessary utility.  I am not comfortable having my datacenter services all fail because a single provider failed.  Having a more expensive location that allows 2+ providers adds to my overhead, but having a backup provider creates peace of mind that any cost savings due to only one provider will not.

4. Building security is important, but not as important as sourcing trustworthy talent to be inside the datacenter.  I would not put a datacenter anywhere that I could not get local, independent sources of talent that all are from a similar cultural background to myself.  This is not xenophobia, it's just being able to relate on a direct level with the people in charge of my business' most critical assets.  For example, If I insult an admin one day just based on a cultural misunderstanding, then I may inadvertently cause my own security issues.

All other conditions can be summarized like this: take what you think is your ideal location and then take a random warehouse building inside a 50 mile radius of where you are right now.  If you can't answer why you should take the ideal location over the random building with all of the reasons above, then you should seriously consider that random building as a candidate.  After all, if it doesn't fail at the above, then exactly why would you not use that as a datacenter?

 

I am a Sr. Systems Engineer for a major telecommunications company. I have a long family history in the Computer Science field.
icon
Greg Ferro
Sun Dec 6 1:39pm
The idea that people are a similar cultural background isn't viable. For example, the Indian government has legislation that forces data centers to be located there. China has more stringent rules.

You will not do business in these countries unless you comply, and you will not be able to resource the people you talk about.

While I can see your point of view, and agree to some extent, it doesn't work that way in my experience.

While it’s easy to understand many elements of Data Center location, there are always hidden aspects that you find out about later. I once worked with a Data Center that was quite safe in most respects : power, cooling, physical security of the building, ownership was solid and so on. We were quite happy with how things were. Except that over the next year there were two problems that really had nothing to do with the Data Center itself or even the organisation that owned it.

The first problem was flooding. Not the facility itself, but the surrounding area. We never thought to check flood access to the building location and when it rained reasonably heavily, the streets, trains and buses that gave our engineers access would be inundated and no access was possible. Of course, this wasn’t a problem until the day that we were experiencing a major outage and no one could get to the data centre to perform the hands on. It only takes a few days without backup tape collection, or new products, or delivery of spare parts to have a real impact on your service levels.

The second problem was the physical safety of the surrounding area. We are often called upon to work nights and weekends to perform upgrades and installation and while the facility was well patrolled and monitored, but most of us were very uncomfortable with travelling to the site at night. The area was well known for criminal activity, and, because it was an industrial area, few people were around to notice if anything happened. Scheduling work at the site was sometimes more complicated because of this.

While the data centre was protected (up to a point at least), lack of access can be a real difficulty. It’s worth taking some time to consider the area AROUND the your facility as well as the facility itself.

Greg Ferro is a freelance Network Architect for large organisations specializing in Data Centre and Security design and operations. A focus on operational outcomes, advanced technologies and creative thinking has proven valuable for customers.

The process of building the physical data center is intensive and specialist. Designers are available who have experience in the building design, the environmental engineer can handle the power, cooling and security and the Planning consultants can handle the regulatory hurdles with planning and approvals and so on. But one area that is often overlooked or given a low priority is the supply of bandwidth.

Many new build Data Centres are placed in geographically remote locations to take advantage of available and hopefully cheap power, good workforces and safe locations. But this remoteness also means that bandwidth will be hard to source.

To ensure that you are getting a fair price, you need at least three, and preferably more, telecommunications providers to compete for your services. Many companies feel that two is sufficient but this always proves false. Your telco may have to lay a long haul of fibre across the countryside to get to your building, and the backbone bandwidth to major POP and Interconnects may also need upgrades which can include additional cabling costs. Cabling projects take months to complete and require large capital outlays by the telco.

From a business perspective, you are only likely to get interest from the telco to make that investment at build time and they can plan to recover the cost from your services over time. Certainly having the bandwidth installed after the data center is complete puts you at a significant disadvantage in the negotiation process and you can expect to pay all capital costs for the installation.

Having multiple suppliers is also important for service redundancy, but mostly vital for negotiating competitive pricing for bandwidth. Because telco costs are typically 5% of the operational expense of a data center, you should ensure that it is a key part of your data centre selection process.

Greg Ferro is a freelance Network Architect for large organisations specializing in Data Centre and Security design and operations. A focus on operational outcomes, advanced technologies and creative thinking has proven valuable for customers.

For starters, you do not choose one location, you choose at least two - and one of those should be a standby location, so you actually need three.

A data center today has two requirements: Electricity and Internet. If you have plenty of that, you could place it anywhere on the planet - the Earth is not big enough for delays in cables to matter to data-center-related communication. To optimize operations you want two independent providers of both network and electricity, which limits the locations a bit more. Cheap bandwidth is also not a rich commodity in places with cheap electricity. And you want a guaranteed supply.

Since cheap electricity and cheap Internet only intersect in a few place, to choose from among those, there are two more factors which are likely to be determining the location, once you have chosen among the places where Internet and electricity is cheap and plentiful. That is facilities and staff. Ideally, you would select a self-cooling location, like Greenland. But that would be a problem from network access and staffing point of view. And a data center in Alaska probably does not save that much in energy for cooling over a data center in California, anyway.

However small data centers have become, they need floor space in a secure building. Which, if the data center was economical in the first place, has to be fairly big because data centers have to be big to be profitable nowadays.  Co-location is an easy solution, but the problem is the same: Big and secure building, good road access. And then, there is staff.

Not that you need very competent security guards, janitors and so on (they just have to know how to clean and not touch any cables), but you need some operations staff. Even if that staff is supported by a remote operations center,  they have to know what to do if something happens, and that is not easy. You do not need a Ph.D. to do the job, but you definitely have to be an engineer. And engineers do not like to live in really boring locations any more than anyone else.

That probably still means there are quite a few places in the midwest of the US and other similar locations that would work. But why just look in the US? Labor costs are much lower in other countries, and even if they are tropical, you could use water for cooling (to heat the pool). Having a data center in Bahamas would be a good perk for the staff. Not to speak of European countries, and some places in Asia - Japan, when the yen becomes cheaper. But the Japanese islands of Okinawa - where the US bases mostly are - have a fairly depressed economy, so there may be incentives in setting up there. India and China, once they get better Internet connectivity, would also work.

Which brings me to a last factor: Government incentives. In many countries, jobs are hard to come by, and while they would not buy the computers, local and national governments in many places would gladly provide facilities for free, if they got some jobs out of it. Sweden comes to mind - they are closing car factories which would work perfectly, and there will be incentives for creating jobs.

Because nothing says that a data center will stay where it is set up. Data centers in containers may still be a bit extreme, but with the pace of computer development and the ease of moving, abandoning the data center after three years, when it is time to invest in a new one, may actually be an option. Just go where it is cheapest then. Which means landlords had better be careful.

When locating a data center, your biggest problem will follow one rule: It's always going to be something that you didn't anticipate. Obviously many companies don't have the luxury of constructing a new data center building from scratch, so instead, they first look around for the cheapest facility that won't break their department's budget - or they're offered an unused piece of in-house real estate. At that point your options can be limited to "choosing the right corner". And your first problem will inevitably be something that's unique to your specific location.

I once worked for a large corporation where I learned the data center's colorful history. Our department was offered the use of a giant warehouse that had originally housed an assembly line. The assembly line was still bolted to the floor and couldn't be removed, so our overworked facilities manager just installed a ring of cubicles around it. And since the servers had to be behind a locked door, our original data center ended up in a wire cage in the corner of the building.

The first problem you'd run into could be as simple as: ventilation. With too much sunlight, your cooling costs could increase, or your performance could decrease. (In one notorious incident, our servers got so hot that the plastic base of the fan started melting!) In the end we were grateful for something we hadn't even planned: our data center was close to the apartment of our systems administrator. Eventually he even installed a thermometer - and then hooked it to a 24-hour web cam!

A better solution might've been to simply choose a facility that had a better cooling system, and ventilation. But ultimately, you should determine the location of a data center using the same priorities as for any other data-related question. Which choice offers the most security, the best redundancy, and the best possible performance?

And then be ready to tweak the hell out of it as soon as something unexpected happens.

David Cassel is a Silicon Valley-based journalist who has been reporting on technology since 1994.