I have an ODL controller running with 198 nodes; each node with 104 NodeConnectors.

My code queries each of the 198 nodes for the nodeConnectors using the following code:

public static List<NodeConnector> getNodeConnectors(DataProviderService dps, String nodeId) {

// Create a Type-safe reader so that casting isn't necessary.
TypeSafeDataReader reader = TypeSafeDataReader.forReader(dps);

// Create the inventory node Id key:
NodeId inventoryNodeId = new NodeId(nodeId);
NodeKey inventoryNodeKey = new NodeKey(inventoryNodeId);

// Get the child node Id:
InstanceIdentifierBuilder<Node> nodeChild = InstanceIdentifier.builder(Nodes.class).child(Node.class,
inventoryNodeKey);
InstanceIdentifier<Node> nodePath = nodeChild.toInstance();

// Read the node from the Operational Data store:
Node node = reader.readOperationalData(nodePath);

if (node == null)

return node.getNodeConnector();
}

This takes about 12 seconds to run on a standard Windows desktop.

An Analysis of the CPU performance using JProfiler shows a very high number of calls to SchemaAwareDataStoreAdapter$MergeFirstLevelReader$2.apply, with the bulk of CPU time spent in this method (specifically Map.equals).

See attachment for screen snapshot of the JProfiler analysis.

For any questions, feel free to contact.