7 Comments

Summary:

Instagram is pretty proud of the infrastructure it built atop the Amazon Web Services cloud, but does an acquisition by Facebook mean goodbye to Amazon? If I were a betting man, I’d say there are already some engineers working really hard to make that happen.

goodbye

Social photo-sharing service Instagram is pretty proud of the infrastructure it built atop the Amazon Web Services cloud, but I have to wonder whether the startup’s acquisition by Facebook today means all that effort was for nothing. Well, it wasn’t for nothing, obviously — Instagram was able to scale to handle tens of millions of users without developing a reputation for being slow or unavailable — but Facebook is pretty adamant about running its services in its own custom-built data centers. And for good reason.

AWS is great for startups with lean budgets, but it’s not unheard for them to leave it behind when they strike it rich. There are plenty of reasons not to remain in the cloud, ranging from the cost to availability to the complexity of building scalable, high-performance web applications on virtual infrastructure. As Facebook explained in its S-1 filing earlier this year, providing an always-on, always-optimized user experience is critical to the success of its platform.

The Instagram engineering team has opened up lately about its core technology stack and how it handled millions of new Android downloads in a single day without going down, but Facebook has been up to slightly bigger things. It has designed its own data centers from the servers up to the facility. It operates one of the world’s largest Hadoop clusters, and very likely the world’s largest MySQL implementation. It has developed everything from a PHP-optimization platform to a NoSQL database to a tool for auto-provisioning and configuring tens of thousands of servers. Its former engineers get acqui-hired to help companies such as Dropbox take their platforms to the next level.

If Facebook considers Instagram a critical part of its mobile strategy going forward — and at a billion-dollar price tag, it probably should — it might not be too keen on letting Instagram keep running on the fairly reliable, but far from perfect AWS cloud. Not when Facebook has all the money and all the talent to make sure Instagram stays up and running safely on Facebook’s own servers.

Facebook has declined to comment on its plans for Instagram’s backend, so this is all just speculation for now. But if I were a betting man, I’d say there are already some engineers working really hard on re-architecting Instagram for the physical world.

Image courtesy of Celeste Hutchins.

  1. Facebook demonstrates fake power with Instagram purchase…
    http://dobrisratings.com/index.php?option=com_content&task=view&id=149532&Itemid=111

    Share
  2. Hey

    Share
  3. What I meant is that it would make sense they go route, I would think. *Sorry for that previous comment.

    Share
  4. Reblogged this on Virtualized Geek Blog and commented:
    This reminds me when Google brought Youtube and migrated it’s services from Rackspace. Wondering how this will affect smaller cloud providers outside of AWS. I believe Amazon will find plenty of companies to utilize the space given up by Instagram but I don’t know how smaller cloud players plan on competing with Amazon in the long run. Specifically Rackspace, as OpenStack stutters with the loss of Citrix does Rackspace have enough runway to get to critical size. Can it remain a standalone company?

    Share
    1. Derrick Harris Monday, April 9, 2012

      I think it’s still too early to tell what will happen with OpenStack, but for the meantime Rackspace looks to be in fine position. The only other (non-AWS and RAX) provider I know of with a good number of these type Web 2.0/3.0 that could get bought is SoftLayer.

      Share
      1. I’ll have to look at SoftLayer. I really like RAX. Their launch of OpenStack is brave. With cable and phone companies buying all the other players (Navisite, Terremark) I fully expected someone to make a play for RAX.

        Share
  5. Alexander Haislip Monday, April 9, 2012

    Really highlights the need to architect for portability, from cloud to cloud and from cloud to on-premise.

    Share

Comments have been disabled for this post