Quantcast: Lessons Learned Rebuilding a Large Scale Real-time Feature Store

Kristi Tsukida, Sr. Software Engineer, Quantcast

My name is Kristi Tsukida, I am a staff engineer. I work at Quantcast and Quantcast is an Ad tech company. We work on targeting ads in real time. We’re just getting started with Aerospike.

We’re using Aerospike to replace one of our legacy systems which help with storage features so that we could know about a user’s past history when we try and figure out which ads we want to be showing.

Limitations on our legacy system were that it wasn’t very flexible for us to be able to change what features are being stored and how we were featurizing and we didn’t understand how to operate it with preventing us from being able to come up with better products to better serve our customers.

I think they’re going to be a lot of benefits to Quantcast as we’re starting to run Aerospike. We’re going to be able to make more datasets available at bid time so that we can make better decisions about who to target ads to and which sites are valuable to different marketers.

For us, Aerospike has been a great way to get up and running with a scalable, maintainable key value store on that we didn’t have to write ourselves. The most surprising thing about Aerospike is probably how easy it was that it didn’t take a whole lot to get up and running and has been very reliable in our tests so far.

Aerospike has been really amazing to work with. We’ve received really good support from everybody that we’ve talked with and even emailing the support team has been, has been an amazingly pleasant experience.