3 ms·
I find a lot of these frameworks to be overkill. A rating event comes into our web server (python), then sent to SQS, then the event is pulled by our custom ar
by poorman 10y ago
I find a lot of these frameworks to be overkill.
A rating event comes into our web server (python), then sent to SQS, then the event is pulled by our custom artificial neural network (deep learning) script, --written in Go. A model is trained and serialized. Next, the serialized model is uploaded to Postgres where it is fetched by the web service (also written in Go) to serve predictions.
We update our models within 15 seconds of a user rating. Every month with millions of ratings, we re-train millions of models and serve billions of predictions.
- achompas 10y agoWhat frameworks are you thinking of, specifically?
- devbug 10y agoInteresting. Do you mind going into more detail? I'm thinking of implementing a similar system – though not using Go.
- poorman 10y agoWhat sort of details are you looking for?
- devbug 10y agoMostly persistence and scaling. I assume you're using a jsonb column, and going horizontal?
- poorman 10y agoHonestly, we haven't needed to scale horizontally. The Postgres column is a jsonb, although probably not necessary since we don't need to query the serialized model. If our reads ever start to slow down, the plan is to just add a follower to read from.
- devbug 10y agoWoops. I was referring to the prediction tier.