updating braodcast info

Author: djpmsft

articles/data-factory/data-flow-exists.md

@@ -37,6 +37,14 @@ To create a free-form expression that contains operators other than "and" and "e
 
 ![Exists custom settings](media/data-flow/exists1.png "exists custom")
 
+## Broadcast optimization
+
+![Broadcast Join](media/data-flow/broadcast.png "Broadcast Join")
+
+In joins, lookups and exists transformation, if one or both data streams fit into worker node memory, you can optimize performance by enabling **Broadcasting**. By default, the spark engine will automatically decide whether or not to broadcast one side. To manually choose which side to broadcast, select **Fixed**.
+
+It's not recommended to disable broadcasting via the **Off** option unless your joins are running into timeout errors.
+
 ## Data flow script
 
 ### Syntax
@@ -46,7 +54,7 @@ To create a free-form expression that contains operators other than "and" and "e
     exists(
         <conditionalExpression>,
         negate: { true | false },
-        broadcast: {'none' | 'left' | 'right' | 'both'}
+        broadcast: { 'auto' | 'left' | 'right' | 'both' | 'off' }
     ) ~> <existsTransformationName>
 ```
 
@@ -65,7 +73,7 @@ NameNorm2, TypeConversions
     exists(
         NameNorm2@EmpID == TypeConversions@EmpID && NameNorm2@Region == DimEmployees@Region,
 	    negate:false,
-	    broadcast: 'none'
+	    broadcast: 'auto'
     ) ~> checkForChanges
 ```
 

articles/data-factory/data-flow-join.md

@@ -64,7 +64,9 @@ Unlike merge join in tools like SSIS, the join transformation isn't a mandatory
 
 ![Join Transformation optimize](media/data-flow/joinoptimize.png "Join Optimization")
 
-If one or both of the data streams fit into worker node memory, further optimize your performance by enabling **Broadcast** in the optimize tab. You can also repartition your data on the join operation so that it fits better into memory per worker.
+In joins, lookups and exists transformation, if one or both data streams fit into worker node memory, you can optimize performance by enabling **Broadcasting**. By default, the spark engine will automatically decide whether or not to broadcast one side. To manually choose which side to broadcast, select **Fixed**.
+
+It's not recommended to disable broadcasting via the **Off** option unless your joins are running into timeout errors.
 
 ## Self-Join
 
@@ -85,7 +87,7 @@ When testing the join transformations with data preview in debug mode, use a sma
     join(
         <conditionalExpression>,
         joinType: { 'inner'> | 'outer' | 'left_outer' | 'right_outer' | 'cross' }
-        broadcast: { 'none' | 'left' | 'right' | 'both' }
+        broadcast: { 'auto' | 'left' | 'right' | 'both' | 'off' }
     ) ~> <joinTransformationName>
 ```
 

articles/data-factory/data-flow-lookup.md

@@ -50,11 +50,11 @@ When testing the lookup transformation with data preview in debug mode, use a sm
 
 ## Broadcast optimization
 
-In Azure Data Factory mapping data flows execute in scaled-out Spark environments. If your dataset can fit into worker node memory space, your lookup performance can be optimized by enabling broadcasting.
-
 ![Broadcast Join](media/data-flow/broadcast.png "Broadcast Join")
 
-Enabling broadcasting pushes the entire dataset into memory. For smaller datasets containing only a few thousand rows, broadcasting can greatly improve your lookup performance. For large datasets, this option can lead to an out of memory exception.
+In joins, lookups and exists transformation, if one or both data streams fit into worker node memory, you can optimize performance by enabling **Broadcasting**. By default, the spark engine will automatically decide whether or not to broadcast one side. To manually choose which side to broadcast, select **Fixed**.
+
+It's not recommended to disable broadcasting via the **Off** option unless your joins are running into timeout errors.
 
 ## Data flow script
 
@@ -67,7 +67,7 @@ Enabling broadcasting pushes the entire dataset into memory. For smaller dataset
         multiple: { true | false },
         pickup: { 'first' | 'last' | 'any' },  ## Only required if false is selected for multiple
         { desc | asc }( <sortColumn>, { true | false }), ## Only required if 'first' or 'last' is selected. true/false determines whether to put nulls first
-        broadcast: { 'none' | 'left' | 'right' | 'both' }
+        broadcast: { 'auto' | 'left' | 'right' | 'both' | 'off' }
     ) ~> <lookupTransformationName>
 ```
 ### Example
@@ -81,7 +81,7 @@ SQLProducts, DimProd lookup(ProductID == ProductKey,
     multiple: false,
     pickup: 'first',
     asc(ProductKey, true),
-    broadcast: 'none')~> LookupKeys
+    broadcast: 'auto')~> LookupKeys
 ```
 ## 
 Next steps