You work for an online travel agency that also sells advertising placements on its website to other companies. You have been asked to predict the most relevant web banner that a user should see…

Question

Options

• AEmbed the client on the website, and then deploy the model on AI Platform Prediction.
• BEmbed the client on the website, deploy the gateway on App Engine, deploy the database on
• CEmbed the client on the website, deploy the gateway on App Engine, deploy the database on
• DEmbed the client on the website, deploy the gateway on App Engine, deploy the database on

• B
  9% (2)
• C
  86% (19)
• D
  5% (1)

Explanation

The requirements are: low latency (300ms at p99), thousands of banners, security, and the simplest solution. Because the inventory of banners is in the thousands and navigation context is a good predictor, the architecture should pre-compute or quickly look up candidate banners. Deploying directly from the client (A) bypasses security controls. The correct architecture (C) uses a gateway on App Engine to handle security and routing, a fast low-latency database (Cloud Bigtable is ideal for millisecond-level key-value lookups at scale) to store pre-computed predictions or banner metadata, and AI Platform Prediction to serve the model. This satisfies the latency SLA, centralizes security at the gateway, and remains simpler than more complex streaming or microservice approaches.

No community discussion yet for this question.