liveBook · Manning

MonitoringService

As explained in Machine Learning Platform Engineering

MonitoringService is a BentoML service component that maintains a window of observed data and generates monitoring reports when configured conditions are met.

MonitoringService supports production drift detection by accumulating API observations into microbatches instead of testing every individual request. section 6.2.3

A monitoring window handles each incoming batch by:

  • Triggering a report when the configured window size is reached or exceeded, then adding it to the workspace. section 6.2.3
  • Evicting older rows so the window contains only the most recent observations. section 6.2.3

Income classifier API

For a /predict endpoint with a window size of 100, the service compares the latest 100 observations with a reference dataset and records each resulting report. section 6.2.3

How Machine Learning Platform Engineering covers it

How deep it goes
Explains initialization, state accumulation, sliding-window eviction, and report triggering, while deferring statistical test details.
Examples
Uses an income classifier, a /predict endpoint, 100-observation windows, and Evidently dashboard snapshots.
What it assumes
Assumes familiarity with API prediction traffic and reference datasets while demonstrating microbatch monitoring from its operational rationale.
What you can do after
Operate/configure BentoML monitoring with a reference dataset, window size, report structure, workspace, and project identifier.
How it connects
Connects microbatching, data drift detection, reference datasets, BentoML services, MLflow, and remote workspaces.
Cover of Machine Learning Platform Engineering
liveBookmanning.com
Read Machine Learning Platform Engineering on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machine Learning Platform Engineering. The section links point to the parts of the book it draws on.