Rec. ITU-T Y.3179 (04/2021) Architectural framework for machine learning model serving in future networks including IMT-2020
Summary
History
FOREWORD
Table of Contents
1 Scope
2 References
3 Definitions
     3.1 Terms defined elsewhere
     3.2 Terms defined in this Recommendation
4 Abbreviations and acronyms
5 Conventions
6 Introduction
7 High-level requirements
     7.1 Inference optimization
     7.2 Model deployment
     7.3 Model inference
8 High-level architecture
     8.1 Architectural components for ML model serving
          8.1.1 Model repository
          8.1.2 ML model serving subsystem
               8.1.2.1 Model optimizer
               8.1.2.2 Inference engine builder
          8.1.3 Inference engine
          8.1.4 MLFO
     8.2 Deployment options
     8.3 High level architecture
     8.4 Reference points
          8.4.1 Reference point 3 between ML sandbox subsystem and ML pipeline subsystem
               8.4.1.1 Serving_model API
          8.4.2 Reference point 5: Interface between management subsystem and ML pipeline subsystem
               8.4.2.1 Model_deployment API
               8.4.2.2 Model_registration API
               8.4.2.3 Serving_model_management API
               8.4.2.4 Model_monitoring_subscription API
               8.4.2.5 Model_monitoring_event API
               8.4.2.6 Health_check API
          8.4.3 Reference point 6: Interface between management subsystem and ML sandbox subsystem
               8.4.3.1 Model_monitoring_subscription API
               8.4.3.2 Model_monitoring_event_notification API
               8.4.3.3 Model_deployment API
          8.4.4 Reference point 16
               8.4.4.1 Model_Push API
          8.4.5 Reference point 17
               8.4.5.1 Serving_model API
               8.4.5.2 Model_push API
          8.4.6 Reference point 18
               8.4.6.1 Inference API
          8.4.7 Reference point 19
               8.4.7.1 Model_optimization API
     8.5 Sequence diagrams of the serving of ML models
          8.5.1 Model optimization in ML model serving subsystem
          8.5.2 Model deployment in ML sandbox subsystem for evaluation
          8.5.3 Model deployment from ML sandbox subsystem to ML pipeline subsystem
          8.5.4 ML pipeline configuration in ML pipeline subsystem
          8.5.5 Model monitoring
          8.5.6 Serving model update
9 Security considerations
Bibliography
<\pre>