liveBook · Manning

Scale

As explained in Software Engineering for Data Scientists

Scale is the size of a data or computational problem, commonly characterized by its number of observations, columns, memory requirements, or efficiency demands.

Scale is the ability of a data or machine-learning workflow to handle growing demand without becoming impractical. It can be understood through memory use and efficiency: ML workloads may consume substantial CPU or GPU time as well as memory. section 1.2.8

A model serving millions of users or processing larger, more diverse datasets may need engineering changes beyond a prototype. In data science, scaling code can involve parallelizing work or choosing advanced data structures so workflows can handle larger data demands. section 1.2.8

Scaling is a software-engineering concern alongside reliability and monitoring. Model retraining frequency depends on performance over time, while monitoring checks that the model continues to meet expected standards. section 1.2.8

How Software Engineering for Data Scientists covers it

How deep it goes
Defines scale through memory and efficiency, naming parallelization and data structures while deferring detailed approaches.
Examples
Uses a model serving millions of users and growing, diverse machine-learning datasets as motivating cases.
What you can do after
Evaluate/understand when an ML workflow needs engineering for memory and efficiency demands.
How it connects
Connects scaling to memory, efficiency, CPU, GPU, parallelization, data structures, monitoring, and retraining.

Also explained in

Compare how 2 books explain scale →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Software Engineering for Data Scientists. The section links point to the parts of the book it draws on.