docs(api): Add additional docs to describe behaviour of model redeployment

docs/user-guide/model_deployment_serving.md

@@ -7,3 +7,19 @@ There are two types of Model Version deployment, standard and python function (P
 Model serving is the next step of model deployment. After we have a running Model Version Endpoint, we can start serving the HTTP traffic by routing the Model Endpoint to it.
 
 ![Model Deployment and Serving](../diagrams/model_deployment_serving.drawio.svg)
+
+# Model Versions and Deployments
+Each model version can deployed with a different set of deployment configurations, such as the number of 
+replicas, CPU/memory requests, autoscaling policy, environment variables, etc. Each set of these configurations that are
+used to deploy a model version are called a *deployment*. 
+
+While each model can have up to **2** model versions deployed at any point of time, each model version can only be 
+deployed using **1** deployment at any point of time. 
+
+Whenever a running model version is redeployed, a new *deployment* is created and the Merlin API server attempts to 
+deploy it, all while keep the existing deployment running. 
+
+If the deployment of the new configuration fails, **the old deployment stays deployed** and remains as the current 
+*deployment* of the model version. The new configuration will then show a 'Failed' status.
+
+![Unsuccessful redeployment](../images/redeploy_model_unsuccessful.png)