AOL's targeted advertising business: Powered by MongoDB

While AOL may evoke thoughts of dial-up broadband for some, the company today drives over $2 billion in annual revenues connecting advertisers to consumers of its premium content, including Huffington Post, Moviefone, Engadget, TechCrunch, Patch, and Stylelist. MongoDB provides the data infrastructure for a significant portion of AOL’s business, both on the content and advertising sides of AOL.

In the words of Jonathan Reed, formerly a senior software engineer at AOL, “AOL uses MongoDB a lot throughout our business,” and for very different use cases. As of June 2012 AOL had over 30 MongoDB projects running internally across over 500 servers.

One of the important projects for which AOL uses MongoDB is advertising, as detailed in the video above. AOL’s Advertising.com platform helps advertisers reach highly-targeted audiences at scale, and MongoDB plays an essential role in storing Advertising.com’s user profiles.

AOL turned to MongoDB for its flexible data model, as user profiles have various sizes and shapes, with different kinds of information stored for different users. One key feature that MongoDB offers is geospatial indexing, which enables AOL to advertise services based on a user’s location (e.g., showing airfare pricing based on the airport nearest to the user, even if all they’ve expressed is interest in flying to Paris).

Importantly, all of this must be done in under five milliseconds, which means that AOL simply can’t afford to hit disk and must keep everything in RAM. MongoDB handles this easily, processing 12,000 transactions per second, or several billion each month. MongoDB’s performance was so good, as Reed describes, the company needed a special set-up to manage network traffic, which couldn’t keep up with MongoDB.

While this seems like it must require a complex set-up, Reed suggests that MongoDB is “surprisingly simple” to set up and run. In the case of Advertising.com for this project, MongoDB runs in a single cluster spanning three data centers, two in the U.S. and one in Europe.

Indeed, ease of use was one of the top-four reasons AOL chose MongoDB to power Advertising.com:

  1. Easy to learn and set up
  2. Easy to scale
  3. Great community
  4. Support contract available (“really good value for money”)

None of this would matter, however, if MongoDB couldn’t handle AOL’s core requirement: dynamic data schema. AOL’s Advertising.com must constantly tweak the kind of user information it collects and stores, and has to be able to do so with super-high performance at scale. MongoDB ticks each of these boxes, and makes it easy to do so, leading Reed to conclude that hitting AOL’s scale requirements “would have been much harder with other technology.”

Tagged with: AOL, Advertising, case study, use cases, flexibility, dynamic schema, high performance, scalability