Redshift Cluster - what type of node Leader is?

In an Amazon Redshift cluster, the leader node type is not explicitly specified or chosen separately from the compute nodes. When you set up a Redshift cluster, you specify the node type (such as RA3 or DC2) and this applies to the compute nodes. The leader node is automatically provisioned as part of the cluster, but its specifications are not directly tied to the compute node type you select.

The leader node manages communications with client applications, parses queries, develops query execution plans, and coordinates the parallel execution of these plans with the compute nodes. It also aggregates intermediate results from compute nodes before returning final results to client applications.

Unlike compute nodes, where you can specify the number and type (like your example of 3 RA3.xlplus nodes), the leader node is provided automatically as part of the cluster architecture. The documentation doesn't specify that the leader node is of the same type as your compute nodes.

The leader node's resources are managed by Amazon Redshift to handle its coordination responsibilities, regardless of the compute node type you choose for your cluster.
Sources
Amazon Redshift provisioned clusters - Amazon Redshift
Data warehouse system architecture - Amazon Redshift
Architecture components of an Amazon Redshift data warehouse - AWS Prescriptive Guidance

Based on the AWS documentation, in an Amazon Redshift cluster, the Leader node is handled differently than the compute nodes.

When you set up a Redshift cluster with 3 ra3.xlplus compute nodes, here's what happens with the Leader node: Leader Node Configuration:

• The Leader node is automatically provisioned and managed by Amazon Redshift
• You do not directly choose or specify the Leader node type
• Its specifications are determined by Amazon Redshift based on your overall cluster configuration

Leader vs. Compute Nodes:

• While your compute nodes are all ra3.xlplus in your example, the Leader node is not necessarily the same type
• The Leader node's resources are automatically scaled by AWS to match the needs of your cluster
• The exact specifications of the Leader node are not directly visible or configurable by users

So to directly answer your question: In a cluster with 3 ra3.xlplus compute nodes, the Leader node is not necessarily an ra3.xlplus node. Its type and specifications are managed internally by AWS and are not explicitly exposed to users. The Leader node is provisioned with appropriate resources to efficiently manage your specified compute nodes.

This approach allows Amazon Redshift to optimize the Leader node's capabilities based on the specific requirements of your cluster configuration without requiring you to make these technical decisions.

We currently use 2 compute nodes of ra3.xlplus. Their utilization revolves around 30% during peak times.

The reason why I asked this question is because we are having issues with Leader Node, which is our current bottleneck. High CPU usage slows things down for users who are running a lot of queries throughout the day.

We applied few "fixes" based on:

https://www.youtube.com/watch?v=y0YHG95MI48&ab_channel=AmazonWebServices

https://www.youtube.com/watch?v=13iIj34nkQE&ab_channel=AWSEvents

https://repost.aws/knowledge-center/redshift-high-cpu-usage

https://repost.aws/knowledge-center/redshift-high-cpu

https://stackoverflow.com/a/70217381

https://aws.amazon.com/blogs/big-data/top-10-performance-tuning-techniques-for-amazon-redshift/

https://docs.aws.amazon.com/redshift/latest/dg/c_SQL_functions_leader_node_only.html

But Leader Node still can't keep up. Our idea was to scale up the cluster, but if Leader is not dependent on that, it does not make too much sense.

I would love to hear how can we increase Leader Node performance.