The concern is documenting the terminology used in the grafana panels such as


1. notes

  • I know WLM === workload management, an interface for setting and configuring job/query priorities

2. leader

  • the basic division of roles in Redshift is
    • leader node
    • compute nodes
  • similar to other services (e.g. master and data nodes in ElasticSearch)
  • its functions:
    1. coordinates the compute notes
    2. handles external communication → applications interact directly only with the leader
    3. develops execution plans
      • mainly series of steps necessary to obtain results for complex queries
    4. compiles code
    5. distributes the compiled code to the copute nodes
    6. allocates a portion of the data to each compute node
  • note that SQL statements are distributed by compute node only when a query references tables stored on a compute node
  • all other queries run exclusively on the leader node
  • some functions can run only on the leader and the query will fail if references tables reside on the compute nodes
    1. leader aggregates intermediary results returned by the compute

3. service class ID

Amazon Redshift WLM creates query queues at runtime according to service classes, which define the configuration parameters for various types of queues

  • a SERVICE CLASS is a range within which a type of queue.
  • each queue has a numerical ID and based on the range it falls into, it is a certain type of the queue
ID Service class
1–4 Reserved for system use.
5 Used by the superuser queue.
6–13 Used by manual WLM queues that are defined in the WLM configuration.
14 Used by short query acceleration.
15 Reserved for maintenance activities run by Amazon Redshift.
100–107 Used by automatic WLM queue when auto_wlm is true.

4. sources