Location, location, location

Much as we might like to think that the Internet exists somewhere out there in the proverbial cloud, there is a real physical infrastructure underlying everything on the Net.  (For a visual tour of the journey of a single bit, see this Wired article: http://www.wired.com/magazine/ff_internetplaces/all/)  And sometimes, when things go awry in the physical world, the damage can have a real effect on the Internet.  Case in point: the websites of Gawker, the Huffington Post, BuzzFeed, and MarketWatch were all knocked offline by Hurricane Sandy.  As of the time of this blog post, nearly 24 hours after the “frankenstorm” made landfall, Gawker is still down.
Gawker’s main data center is located in New York City, which raises the question: why would anyone locate a data center there?  In the past decade or so, New York City has seen a major terrorist attack, not to mention several serious weather events, including Hurricanes Irene and Sandy.  Why not locate instead in, say, Kansas?

IO, a data center infrastructure design company, makes it clear that while environmental threats are one item to consider when selecting where to locate a data center, there are several other important criteria as well (http://www.iodatacenters.com/blog/selecting-the-best-data-center-location/):

  • Quality of local utilities
  • Proximity to telecom carriers
  • Skilled employee base
  • Physical space available
  • Taxes

New York City would probably score highly on proximity to telecom carriers and skilled employee base, but very poorly on taxes and the threat of environmental or man-made destruction, which makes me wonder about Gawker’s decision.

Google has chosen to distribute its U.S. data centers across the country (http://www.google.com/about/datacenters/inside/locations/):

The New York Times has an excellent graphic on the risk of natural disasters in the U.S.  Notably (but not surprisingly), Google’s data centers are well-distributed.  It is unlikely that any two would be offline at the same time for the same reason (http://www.nytimes.com/interactive/2011/05/01/weekinreview/01safe.html):

Maybe Gawker should think about relocating to Kansas after all.

Advertisements
This entry was posted in Uncategorized. Bookmark the permalink.

3 Responses to Location, location, location

  1. samargolis says:

    The Huffington Post was down this morning, it is certainly a frustrating problem to deal with.

    Based on those maps, I believe the Tornado risk in Kansas is worse than the Hurricane risk in New York. The best way to do it is to spread datacenters out Google-Style. However, the concern with that is it is quite expensive to maintain spread-out datacenters (Google itself may even have problems as questions are now being raised about its business model).

    These freak disasters just seem to happen no matter where you place this important equipment. I think the only way we can make these problems less prevalent is to reduce the cost of setting up a datacenter so that a company or research enterprise can easily have several datacenters which they can spread out.

  2. kristenmayer says:

    I agree that setting up a datacenter in Kansas would make it more prone to being down due to tornadoes. Rather than Kansas, it might be better to set up somewhere more to the west – perhaps Washington.

    However, regardless of whether an area is more prone to natural disasters, things will still happen, no matter where you are. Because of this, many companies probably weight other factors, such as “skilled employee base” to be more important. After all, while inclement weather might cause trouble for a few days, maybe even weeks a year, not having skilled employees presents a much bigger problem.

    Of course, companies with the means to do so would be best to spread out their datacenters across the country as Google has.

  3. Sabarish says:

    Nicely written article!

    There is a probably a lot of data out there for all the parameters specified by IO (and other datacenter experts) in some form or the other, so maybe all that is required is for someone to write an algorithm to analyze all that data; find patters and correlations and generate a list of ideal locations to have a datacenter. Who knows what that algorithm will come up with? The results might surprise us all or might be obvious. Maybe New York isn’t a bad location or maybe it is Alaska or perhaps even underground? An interesting research project opportunity, if it hasn’t been done so far.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s