In an article from Geekwire, Dropbox announces cost savings by moving away from AWS and towards an in-house solution.
Dropbox was an early adopter of a cloud solution for a revolutionary product. In essence, they took Amazon’s AWS S3 service, which is used to store flat files, and built a product that was easy to use. I’ve been a user and fan of Dropbox for years, and it is still my primary platform for sharing files among all my devices.
So if this was so successful, why would Dropbox decide to build their own infrastructure instead of using AWS?
I would imagine that the biggest reason is that Dropbox can’t optimize costs using S3. S3 offers many different categories for storage of files. However, Dropbox files need to always maintain the same reliability and need to be readily available for users.
Typically, in an AWS implementation, S3 file strategy around archival is part of the equation. Files become cheaper as you move them along the archival process.
So Dropbox cannot take advantage of such features like Glacier, which allows you to store inactive files and retrieve them within a couple hours time after request. This would be a deal breaker for most Dropbox users.
It also can’t really used IA, which is Infrequent Access, meaning you can store a file normally for cheaper if you rarely expect to need it. It’s instant access, but basically costs more if you assess it frequently, as access is supposed to be “infrequent”. So that storage wouldn’t work for Dropbox users.
Another option is to store it RRS (Reduced Redundancy Storage), but that heightens the chance you will lose the file, as it doesn’t guarantee the 99.999999999% durability you normally get with a file in standard S3. Instead you get 99.99% durability, which means that out of every 10k files, you’ll lose one.
So because Dropbox needed the most expensive option that S3 offered for all tiles, and because files are what Dropbox is all about, it made sense to build their own datacenter to accommodate their explicit needs.
Also, one wonders if the news in the press about S3 buckets being constantly scanned for access and security vulnerabilities played any role in Dropbox wanting to move their service outside of AWS. Perhaps Dropbox, which has their own security model, thinks they will do a better job than Amazon.
There are some concerns that I think Dropbox will need to evaluate in their product though as they move to private data centers.
First, will Dropbox be able to provide fast access to files (for reading, writing, and updating) from all places in the world? It looks like Dropbox’s datacenters are located in the United States, and if one is trying to upload photos from South Korea after attending the Olympics, there may be some delays.
Second, does this limit future features from Dropbox? By building a file-only type service, they may find down the road that they wish to do more in the cloud than just provide file access. If the files were stored in S3, they could leverage any of the AWS 100+ services they offer to build additional functionality. Will Dropbox continue to need to build more features than just an S3-type service?
Third, will Dropbox start to offer their file sharing services (or other services) beyond what they do already? Will they want to scale up to be a full cloud provider like Azure, and provide competition to AWS?
Time will tell, but Dropbox’s decision points out that moves to the cloud are fluid, one can change providers, or decide to re-house their applications, depending on the business need. A provider shouldn’t become complacent in the services they offer – but continue to expand services to provide better functionality over time.