High-Scale AI/ML Feature Serving at Low Cost with Caching

https://www.tecton.ai/blog/high-scale-feature-serving-at-low-cost-with-caching-machine-learning/

by mihirmathur • 3 years ago

24 3 3 years ago

chuckhend• 3 years ago

I haven't used Tecton too much outside of POC, but I have worked with Feast quite a bit. Anyone able to compare/contrast Redis as an online store w/ Feast to Redis cache on Tecton? Are they the same thing?

bsososeo• 3 years ago

They're not the same thing. You can use Redis for either a cache or a durable-ish online store. The cache is purely an ephemeral storage layer. You could in theory bundle up a Redis Cache in front of a Redis Online Store, but it won't help much.

This solution is mainly meant to complement an online store backed with Bigtable or Dynamo to improve response times and cut down on infrastructure costs, especially when using serverless dynamo where you're paying per request.

Generally it's cheaper to store larger volumes of data on Bigtable/Dynamo at the expense of latency and request volume. Redis tends to do really well at low latency, but is fairly expensive if your datasets are large, and becomes somewhat problematic if your dataset is frequently changing and or growing in size. This is primarily due to the difficulty in scaling a Redis cluster up or down in a timely fashion. Using a cache in front of dynamo kind of gets you the best of both worlds.

nvrmnd• 3 years ago

Is this an advertisement?

High-Scale AI/ML Feature Serving at Low Cost with Caching

Related Stories