Using Google Cloud to serve 10,000s of personalized recs per second

In retail, it is personalize or die. Bluecore has a long history using BigQuery to generate and apply personalized recommendations. While this continues to work well in bulk, Google Cloud offers a variety of tools to enable real-time personalization.

These use cases require not only efficient code, but also data storage options that fit real-time access patterns. Products and customers alone present unique challenges that a small team would struggle to address in a self-hosted environment. Keeping data up to date and caches fresh while facing nearly constant updates is always a challenge. Using Google Cloud, Bluecore built a system that delivers tens of thousands of personalized recommendations per second, with a p95 SLO of under 100ms. They did it with a tiny team and in just a few months.

Bluecore touches on how it uses the BigQuery Storage API, Cloud Storage, Bigtable, Datastore, Memorystore, Pub/Sub, Google Kubernetes Engine, and even a little App Engine standard to help retailers activate their customers and meet their revenue goals.

Speakers: Alexa Griffith, Mike Hurwitz

Watch more:
Google Cloud Next ’20: OnAir →

Subscribe to the GCP Channel →


product: App Engine, BigQuery, Cloud Pub/Sub;

Duration: 00:17:49
Publisher: Google Cloud
You can watch this video also at the source.