Add feedback

Author: PatAltimore

articles/iot-operations/manage-mqtt-broker/howto-configure-availability-scale.md

@@ -164,11 +164,11 @@ To learn more, see [`az iot ops create` optional parameters](/cli/azure/iot/ops#
 
 ---
 
-There are predefined memory profiles with different memory usage characteristics.
+There are predefined memory profiles with different memory usage characteristics for publishing messages. There isn't a limit on the number of sessions or subscriptions that the broker can handle. The memory profile governs only the memory usage for PUBLISH traffic.
 
 ### Tiny
 
-Use this profile when you need to handle a small number of connections and messages and have limited memory resources.
+Use this profile when you have limited memory resources and client publish traffic is low.
 
 When you use this profile:
 
@@ -182,7 +182,7 @@ Recommendations when you use this profile:
 
 ### Low
 
-Use this profile when you need to handle a low number of connections and messages.
+Use this profile when you have limited memory resources and clients publish small packets.
 
 When you use this profile:
 
@@ -196,7 +196,7 @@ Recommendations when you use this profile:
 
 ### Medium
 
-Use this profile when you need to handle a moderate number of connections and messages.
+Use this profile when you need to handle a moderate number of client messages.
 
 Medium is the default profile.
 
@@ -205,7 +205,7 @@ Medium is the default profile.
 
 ### High
 
-Use this profile when you need to handle a large number of connections and messages.
+Use this profile when you need to handle a large number of client messages.
 
 - Maximum memory usage of each frontend replica is approximately 4.9 GiB, but the actual maximum memory usage might be higher.
 - Maximum memory usage of each backend replica is approximately 5.8 GiB multiplied by the number of backend workers, but the actual maximum memory usage might be higher.
@@ -237,6 +237,9 @@ In comparison, the *Tiny* memory profile has a frontend memory usage of 1.9 GB a
 
 *2 * 99 MB + (2 * 2) * 102 MB * 2 = 198 MB + 816 MB* = 1.014 GB.
 
+> [!IMPORTANT]
+> The MQTT broker starts rejecting messages when memory is 75% full.
+
 ## Cardinality and Kubernetes resource limits
 
 To prevent resource starvation in the cluster, the broker is configured by default to [request Kubernetes CPU resource limits](https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/). Scaling the number of replicas or workers proportionally increases the CPU resources required. A deployment error is emitted if there are insufficient CPU resources available in the cluster. This notification helps you avoid situations where the requested broker cardinality lacks enough resources to run optimally. It also helps to avoid potential CPU contention and pod evictions.