Aaron is the CTO and co-founder of SproutSocial, a social media management platform for businesses. He told us about some of the issues he encountered while rapidly scaling the tech team and how he used "cloud-backed localStorage" to make development faster.
When SproutSocial first got started, Aaron told us, all the developers were full-stack. There was one single codebase.
After 1.5 years, the system had grown quite complicated, and the tech team needed to split and specialize. SproutSocial divided the team into "front-end" and "back-end" devs. The front-end developers, experts in usability and design, focused on the user experience and user-facing features. The back-end developers, meanwhile, tackled performance and scaling issues.
With the new specializations, though, a bottleneck formed. If a front-end dev wanted to create a new user-facing feature that needed to track some information about a user's choice, he would need to submit a request with the back-end team to add a field to the database and set the field at the appropriate time. This process, Aaron told us, seriously delayed the implementation of new features.
For example, here are some front-end features that would require data persistence:
- Dashboard has advertisements for new features that the user can choose to hide. Need to track which users have hid which features.
- SproutSocial wants to promote its mobile app to each user once; it therefore needs to track which users saw the mobile-app promotion.
- Tracking the user's inbox config preferences.
- Storing which sub-view the user saw last.
Rather than creating many database schemas with sets of CRUD code and read/write APIs for every minor front-end feature, SproutSocial wanted a single consistent mechanism.
Here is a rough overview of how it works:
- The server stores the key/value pairs in a giant MySql table.
"It let us divide our team at a critical time when we needed to divide our team."—Aaron Rankin, CTO of SproutSocial
I asked Aaron how his team did reporting on the massive MySql table. He told me they don't. They have a separate system for reporting on which users did what. They use transparent 1x1 tracking pixels to track which users are doing what, and then they have a Node.js service that sends the tracking pixel data to the server and stores it.
The Node.js-based reporting service is called, in keeping with the plant-based theme, "Pollinator."
Here is a video of Aaron's talk:
Cloud-backed localStorage by Aaron Rankin from Groupon Engineering on Vimeo.