NUMA architecture support

[This documentation is for preview only, and is subject to change in later releases. Blank topics are included as placeholders.]

As hardware capabilities grow and keep growing, software applications need to keep updated to take advantage of all hardware progress. Analysis Services introduced in Denali version of the product support for NUMA configurations and the ability to correctly use more than 64 processors.

The Analysis Services administrator now has the possibility to configure how the different processor groups are to be used and inside those groups which processors used (or prevent some processors from being used).

Thread Pools

Standard task like: Query, Process, IO activity and Formula Engine calculations can now be tied to processor groups by associating the corresponding thread pool to the group.

Affinity Masks

Processors in a processor group can be included or excluded from Analysis Services operations by marking them in the processor group affinity mask.

Memory Allocation

Memory ranges can be defined and assigned to processor groups by a memory list mechanism that allows specifying where to look first for memory requirements.

The following is a list of the most relevant configuration properties that help Analysis Services to support NUMA architectures

  • \VertiPaq\ThreadPool\GroupAffinity

  • \ThreadPool\Process\GroupAffinity

  • \ThreadPool\Process\GroupAffinity

  • \ThreadPool\Query\GroupAffinity

  • \ThreadPool\Parsing\Long\GroupAffinity

  • \ThreadPool\Parsing\Short\GroupAffinity

  • \ThreadPool\IOProcess\GroupAffinity