ML System Architecture

Created: 2023-02-23 12:01
#quicknote

Definition of Architecture: the way software components are arranged and the interactions between them. See Architecture component breakdown for a detailed analysis of the key elements.

There are several elements to consider, as shown in the next image:

The minimal capacity needed to offer the service is fundamental. For this, remember to consider:

system and service load estimates (average versus peak loads);
requirements for a rollout strategy;
autoscaling for optimal use of the system;
overload mitigation strategies

Obviously, the cost planning is also important -> ML services can overrun costs if not properly planned and monitored.

ML System Architecture

Tags