Canary deployment: It rolls out the new ML system to a small fraction of the traffic initially, where the ML system is used in the decision making process. A canary deployment gives you the flexibility to ramp-up how much traffic you divert to your new ML system slowly.
Blue Green deployment: It diverts some or all of the system’s traffic to the new ML system, which is used to make decisions.
Shadow mode deployment: It is a type of deployment where the ML algorithm runs in parallel with the user but it’s output isn’t used for any decision making.