WLM techniques

There are several techniques for managing resource allocations:
Table 1. Workload management feature summary
Technique Description
Scheduler rules Scheduler rules influence the scheduling of plans. Each scheduler rule specifies a condition or set of conditions. Each time the scheduler receives a plan, it evaluates all modifying scheduler rules and carries out the appropriate actions. Each time the scheduler selects a plan for execution, it evaluates all limiting scheduler rules. The plan is executed only if doing so would not exceed a limit imposed by a limiting scheduler rule. Otherwise, the plan waits. This provides you with a way to classify and manipulate plans in a way that influences the other WLM techniques (SQB, GRA, and PQE).
Guaranteed resource allocation (GR) You can assign a minimum share and a maximum percentage of total system resources to entities called resource groups. The scheduler ensures that each resource group receives system resources in proportion to its minimum share. A resource group receives a larger share of resources when other resource groups are idle, but never receives more than its configured maximum percentage. Each plan is associated with a resource group, and the settings of that resource group settings determine what fraction of available system resources are to be made available to process the plan.
Short query bias (SQB) Resources (that is, scheduling “slots”, memory, and preferential queuing) are reserved for short queries. A short query is a query for which the cost estimate is less than a specified maximum value (the default is 2 seconds). With SQB, short queries can run even when the system is busy processing other, longer queries.
Prioritized query execution (PQE) Based on settings that you configure, the system assigns a priority (critical, high, normal, or low) to each query. The priority can be made to depend on factors such as the user, group, or session associated with the query. The system can then use the priority as a basis for allocating resources and scheduling work.
Configuring the WLM to provide predictable performance, eliminate queue wait and disk-based queries
  • Scaling workloads by giving them enough resources (for example, concurrency and memory).
  • Isolating and protecting your predictable workloads (for example, batch operations) from unpredictable workloads (for example, ad hoc queries from reporting tools).
  • Scheduler rules (distribute queries throughout the day), guaranteed resource allocation (GRA – reducing disk-based queries), short query bias (SQB – reduce queue time) and prioritized query execution (PQE - reduce queue time).
  • Use scheduler rules or resource group assignments to classify plans into groups, then use guaranteed resource allocation (GRA) to arrange for each group to receive a different portion of system resources. Within each resource group, you can reserve scheduling and memory resources for short queries (SQB). Within each resource group, use priority query execution (PQE) settings to prioritize more important jobs over less important jobs.

Use any combination of these techniques as required by the methodology that you employ to manage queries.

Important: Work with your Netezza Performance Server sales or support representative to assess which WLM techniques are appropriate for your situation. Changes can affect system behavior in unintended ways and must be carefully planned for and implemented.