Add network topology e2e test cases

[JesseStutler]

Network topology needs to add e2e test cases

The current scoring mechanism of hypernode is as follows:
Score(sum) = Score(hypernodes) + Score(nodes)

• Score(hypernodes): The lower the tier of the Hypernode, the higher the score. If the tier is the same, the more Pods deployed under the vc-job in the hypernode, the higher the score.
• Score(nodes): Sum the score of the Nodes in the HyperNode, and the Node score is calculated according to binpack

Suppose we create the HyperNodes structure according to this pic, and the test cases are also tested according to this pic. (We can create fake nodes using kwok)

Test Cases

• case 1: network topology mode is hard, HighestTierAllowed is 1, assume that there are already Pods occupying resources on Node-0 at this time. Then deploy a vc-job that needs to use NetworkTopology. Because node scoring is based on binpack, we will try to fill the nodes first as much as possible.
  • subcase1-1: The score of hypernode s0 is highest. The resources of node-0 are enough to be allocated to all pods in vc-job. These pods should be scheduled to node-0.
  • subcase1-2: The score of hypernode s0 is highest. The resources of node-0 are not enough to be allocated to all pods in vc-job. Then part of the pods should be scheduled to node-0, and part of the pods should be scheduled to node-1.
  • subcase1-3: Neither Hypernode s0, s1, s2, nor s3 resources are enough to be allocated to vc-job, all/part of the pods of the vc-job should be pending.
• case 2: network topology mode is hard, HighestTierAllowed is 2. Assume that the hypernodes resources of tier 1 are not enough to be allocated to vc-job. Tier 2 hypernode resources are enough to be allocated to vc-job. If there are already some pods occupying in hypernode s4, s4 score should be higher, and all pods should be scheduled to hypernode s4.
• case 3: network topology mode is hard, HighestTierAllowed is 3. Assume that there are already parts of vc-job pods scheduled onto hypernode s4, but there are still pods in vc-job needed to be scheduled, then these pods should be scheduled onto hypernode s4, because more pods deployed under the vc-job in the hypernode, the higher the score.
• case 4: network topology mode is soft, Assume that the hypernodes resources of tier 1 are not enough to be allocated to vc-job, tier 2 hypernode resources are enough to be allocated to vc-job(Need 4 nodes resources). The pods of vc-job should be scheduled onto hypernode s4 or s5, cannot be scheduled both in s4/s5.

[JesseStutler]

/kind feature

[JesseStutler]

/area test

[Xu-Wentao]

/assign