From the course: Advanced Data Processing: Batch, Real-Time, and Cloud Architectures for AI
Join today to access over 24,900 courses taught by industry experts.
Scaling batch AI
From the course: Advanced Data Processing: Batch, Real-Time, and Cloud Architectures for AI
Scaling batch AI
- [Instructor] How do we scale batch AI? Let's discuss the levers and focus areas for scaling batch AI in this video. When it comes to batch AI, scaling needs are determined by the expected data volumes and the latency thresholds for the use case. Higher data volumes and lower latency requirements would mean more scaling is needed for the architecture. What are some of the levers that can be used to scale a batch AI architecture? We begin with scaling feature engineering. Concurrent reads from data sources using multiple connections can read more records in less amount of time. Using big data processing patterns, like MapReduce, can increase horizontal scalability of the pipeline. Using N + 1 deployments for the jobs can provide parallel processing of records. Concurrent writes to data sinks using multiple connections can also increase write throughput. Using on-demand scaling to increase the number of compute resources can help handle spikes in loads. How can we scale model training?…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.
