Identify the Differences and Advantages of Model Serving Approaches: Batch, Realtime, and Streaming MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   An e-commerce site wants to provide product recommendations instantly when a user views a product page. Which serving approach is most appropriate?

   • ABatch serving with a faster schedule
   • BBatch serving
   • CRealtime serving
   • DStreaming serving
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   What is a primary advantage of realtime model serving?

   • AIt provides immediate, low-latency responses to individual requests
   • BIt can process unbounded streams of data continuously
   • CIt is simpler to implement than batch serving
   • DIt is the most cost-effective for processing massive datasets
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   Which trade-off is most commonly associated with streaming serving compared to batch serving?

   • AStreaming is more expensive but provides no latency benefit
   • BStreaming provides better data freshness but adds operational complexity
   • CStreaming is simpler but provides stale data
   • DStreaming is cheaper but has higher latency
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A ride-sharing company wants to predict surge pricing. They currently run a batch job every 15 minutes, but drivers complain that prices are outdated. They need fresher predictions but cannot afford the complexity of a full streaming pipeline. Which approach offers the best trade-off between freshness and complexity?

   • AImplement streaming serving with a 1-second trigger
   • BKeep the 15-minute batch and accept the staleness
   • CSwitch to realtime serving with a REST endpoint per ride request
   • DReduce the batch interval to every 5 minutes
   Reveal Solution Discussion Select an answer first
6. 20expert · hard

   A news website wants to personalize article recommendations. They have two requirements: (1) a user's homepage must load recommendations in under 200 ms, and (2) the recommendation model should incorporate the user's latest reading behavior within 5 minutes. The team is considering streaming serving with a 1-minute trigger versus realtime serving with a precomputed feature store. Which approach best balances the latency and freshness requirements?

   • ARealtime serving with features computed from the last 24 hours of behavior
   • BRealtime serving with a feature store updated by a streaming job
   • CBatch serving that runs every 5 minutes on user behavior
   • DStreaming serving that scores each recommendation as the user loads the page
   Reveal Solution Discussion Select an answer first