Episode 6 of this season of Architecture Corner is out (I made a guest appearance in episode 1, “Good at Innovation”). In this installment, the CIO is a glutton for new data center capacity.
Chris the CEO (Casimir Artmann) and John the CIO (Greger Wikstrand) are convinced that “we are special”. Can Ann the CFO (Christina Lundström) and a cloud computing expert convince them to explore the alternative to an expensive new data center?
One of the big news items from last week was DropBox’s announcement that it had brought its file storage infrastructure in-house, moving (mostly) away from AWS:
Years ago, we called Dropbox a “Magic Pocket” because it was designed to keep all your files in one convenient place. Dropbox has evolved from that simple beginning to become one of the most powerful and ubiquitous collaboration platforms in the world. And when our scale required building our own dedicated storage infrastructure, we named the project “Magic Pocket.” Two and a half years later, we’re excited to announce that we’re now storing and serving over 90% of our users’ data on our custom-built infrastructure.
Given both the massive scope of the endeavor and the massive repudiation of what’s becoming more and more common infrastructure practice, Wired.com‘s article is appropriately titled: “The Epic Story of Dropbox’s Exodus From the Amazon Cloud Empire”. According to that article:
In essence, they built their own Amazon S3—except they tailored their software to their own particular technical problems. “We haven’t built a like-for-like replacement,” (Dropbox engineering VP Aditya) Agarwal says. “We’ve built something that is customized for us.”
Did DropBox make the right decision?
Only time will tell. In truth, it will probably be easier to tell if it was the wrong decision than if it was the right one. Poor choices tend to be more absolute, good choices less clear-cut.
DropBox has also spent more than two years developing and proving their infrastructure accord to their announcement. Given that file storage is DropBox’s core business and recognizing the scale they operate at (500 million users and 500 petabytes of data according to an article on CIO.com), the idea that they should control their own infrastructure makes sense. This is particularly true given Agarwal’s statement that “We’ve built something that is customized for us.” While I can’t say whether the decision is right, I can say that it is a reasonable one.
That’s not the same, however, as saying that everyone should operate the own infrastructure. Context matters. Simon Wardley‘s tweet sums it up nicely (where “nicely” is defined as “snarkily”):
Emulating DropBox when you’re not in the storage business, when you don’t have the volume (nor, most likely, the budget), and when you haven’t done the homework to prove it out, makes no sense. It would be like rowing a dinghy out into a stormy ocean because the oil tanker that just left port is doing fine.
Context matters.
Form Follows Function 