Affinity Scheduling Policies

Table 1 describes the resource usage rules of the Atlas 900 A3 SuperPoD.

**Table 1** Affinity policies
Priority	Policy Name	Policy Description
1	Full node occupation	A node with fewer processors indicates a higher priority. NOTE: When a single-server job is delivered, the job does not fully occupy a node as the full priority scheduling principle required. Description: Symptom: For the Atlas 900 A3 SuperPoD, if two-processor and 14-processor jobs are scheduled at the same time, the two jobs are not scheduled to the same node. Cause analysis: After Volcano schedules a job, there is a delay for Ascend Device Plugin to report the scheduled Ascend AI Processor topology to mindx-dl-deviceinfo-$*{node_name}*. As a result, Volcano fails to verify the number of Ascend AI Processors on a node and the job is scheduled to another node.
2	Preference for remaining nodes	When there are two reserved SuperPoDs, one with three remaining nodes and the other with two remaining nodes, select the SuperPoD which has three remaining nodes.
3	Full occupation of SuperPoDs	When there are two reserved SuperPoDs, one with four remaining nodes and the other with three remaining nodes, select the SuperPoD which has three remaining nodes.

Parent topic: Atlas 900 A3 SuperPoD