Differences between Priority Scheduling in SLES 10 and SLES 11 - Teradata Database - Teradata Viewpoint

Workload Pre-Migration User Guide

Product
Teradata Database
Teradata Viewpoint
Published
January 2016
Language
English (United States)
Last Update
2018-05-07
dita:mapPath
yfc1467307781404.ditamap
dita:ditavalPath
ft:empty
dita:id
scv1467244925615
Product Category
Analytical Ecosystem
Software
Teradata Database manages priority scheduling differently in SLES 10 and SLES 11.
The following descriptions are very basic and designed to highlight the main differences in priority scheduling between SLES 10 and SLES 11. For additional detail on priority scheduling and other aspects of workload management in SLES 11, see Teradata Viewpoint User Guide.

Priority Scheduling in SLES 10

In SLES 10, Teradata Database uses resource partitions, performance groups, performance periods, and allocation groups to schedule priority.

At the highest level, the user can define up to four resource partitions and assigns a weight value to each. This value determines the proportion of resources available to a partition relative to the other partitions. In addition, the user associates each partition with at least one performance group, which, in turn, is associated with between one and eight performance periods, and with at least one allocation group. User queries run in allocation groups based on a shared enforcement priority (EP) of either Tactical, Priority, Normal, or Background. The combination of resource partition and allocation group parameters, acting through their common performance groups, determines the precise priorities of workloads running on Teradata Database.

The following figure illustrates priority scheduling in SLES 10.

Priority Scheduling in SLES 11

In SLES 11, Teradata Database takes a tiered approach to priority scheduling. By default, a single virtual partition is defined at the top level, with up to ten virtual partitions allowed (depending on licensing). Each partition includes one Tactical tier and one Timeshare tier, by default, and (again, depending on licensing) one optional Service Level Goal (SLG) tier, with up to five SLG tiers allowed within a virtual partition. Tiers, in turn, contain workloads. Like an allocation group, a workload represents a group of queries with common characteristics. But unlike allocation groups in SLES 10, workloads in SLES 11 are used directly to determine scheduling priority.

At the broadest level, tier assignment determines priority, with workloads in the Tactical tier taking highest priority, followed by workloads in the SLG tier 1, followed by workloads in SLG tiers 2 through 5 (in order), followed by workloads in the Timeshare tier. At a finer level, a user-configurable allocation percentage for each workload in the SLG tier determines priority within that tier. Likewise, a user-configurable access level for each workload in the Timeshare tier determines priority within that tier as either Top, High, Medium, or Low. In addition to user-specified workloads, the Tactical and SLG tiers also contain a default Remaining workload to ensure that some resources always flow to tiers below.

The following figure illustrates priority scheduling in SLES 11.