Resource Bin Packing for Extended Resources

FEATURE STATE: Kubernetes v1.16 [alpha]

The kube-scheduler can be configured to enable bin packing of resources along with extended resources using RequestedToCapacityRatioResourceAllocation priority function. Priority functions can be used to fine-tune the kube-scheduler as per custom needs.

Enabling Bin Packing using RequestedToCapacityRatioResourceAllocation

Kubernetes allows the users to specify the resources along with weights for each resource to score nodes based on the request to capacity ratio. This allows users to bin pack extended resources by using appropriate parameters and improves the utilization of scarce resources in large clusters. The behavior of the RequestedToCapacityRatioResourceAllocation priority function can be controlled by a configuration option called requestedToCapacityRatioArguments. This argument consists of two parameters shape and resources. The shape parameter allows the user to tune the function as least requested or most requested based on utilization and score values. The resources parameter consists of name of the resource to be considered during scoring and weight specify the weight of each resource.

Below is an example configuration that sets requestedToCapacityRatioArguments to bin packing behavior for extended resources intel.com/foo and intel.com/bar.

  1. apiVersion: v1
  2. kind: Policy
  3. # ...
  4. priorities:
  5.   # ...
  6.   - name: RequestedToCapacityRatioPriority
  7.     weight: 2
  8.     argument:
  9.       requestedToCapacityRatioArguments:
  10.         shape:
  11.           - utilization: 0
  12.             score: 0
  13.           - utilization: 100
  14.             score: 10
  15.         resources:
  16.           - name: intel.com/foo
  17.             weight: 3
  18.           - name: intel.com/bar
  19.             weight: 5

This feature is disabled by default

Tuning the Priority Function

shape is used to specify the behavior of the RequestedToCapacityRatioPriority function.

  1. shape:
  2.  - utilization: 0
  3.    score: 0
  4.  - utilization: 100
  5.    score: 10

The above arguments give the node a score of 0 if utilization is 0% and 10 for utilization 100%, thus enabling bin packing behavior. To enable least requested the score value must be reversed as follows.

  1. shape:
  2.   - utilization: 0
  3.     score: 100
  4.   - utilization: 100
  5.     score: 0

resources is an optional parameter which defaults to:

  1. resources:
  2.   - name: CPU
  3.     weight: 1
  4.   - name: Memory
  5.     weight: 1

It can be used to add extended resources as follows:

  1. resources:
  2.   - name: intel.com/foo
  3.     weight: 5
  4.   - name: CPU
  5.     weight: 3
  6.   - name: Memory
  7.     weight: 1

The weight parameter is optional and is set to 1 if not specified. Also, the weight cannot be set to a negative value.

Node scoring for capacity allocation

This section is intended for those who want to understand the internal details of this feature. Below is an example of how the node score is calculated for a given set of values.

Requested resources:

  1. intel.com/foo : 2
  2. Memory: 256MB
  3. CPU: 2

Resource weights:

  1. intel.com/foo : 5
  2. Memory: 1
  3. CPU: 3

FunctionShapePoint {{0, 0}, {100, 10}}

Node 1 spec:

  1. Available:
  2.   intel.com/foo: 4
  3.   Memory: 1 GB
  4.   CPU: 8
  6. Used:
  7.   intel.com/foo: 1
  8.   Memory: 256MB
  9.   CPU: 1

Node score:

  1. intel.com/foo  = resourceScoringFunction((2+1),4)
  2.                = (100 - ((4-3)*100/4)
  3.                = (100 - 25)
  4.                = 75                       # requested + used = 75% * available
  5.                = rawScoringFunction(75) 
  6.                = 7                        # floor(75/10) 
  8. Memory         = resourceScoringFunction((256+256),1024)
  9.                = (100 -((1024-512)*100/1024))
  10.                = 50                       # requested + used = 50% * available
  11.                = rawScoringFunction(50)
  12.                = 5                        # floor(50/10)
  14. CPU            = resourceScoringFunction((2+1),8)
  15.                = (100 -((8-3)*100/8))
  16.                = 37.5                     # requested + used = 37.5% * available
  17.                = rawScoringFunction(37.5)
  18.                = 3                        # floor(37.5/10)
  20. NodeScore   =  (7 * 5) + (5 * 1) + (3 * 3) / (5 + 1 + 3)
  21.             =  5

Node 2 spec:

  1. Available:
  2.   intel.com/foo: 8
  3.   Memory: 1GB
  4.   CPU: 8
  5. Used:
  6.   intel.com/foo: 2
  7.   Memory: 512MB
  8.   CPU: 6

Node score:

  1. intel.com/foo  = resourceScoringFunction((2+2),8)
  2.                =  (100 - ((8-4)*100/8)
  3.                =  (100 - 50)
  4.                =  50
  5.                =  rawScoringFunction(50)
  6.                = 5
  8. Memory         = resourceScoringFunction((256+512),1024)
  9.                = (100 -((1024-768)*100/1024))
  10.                = 75
  11.                = rawScoringFunction(75)
  12.                = 7
  14. CPU            = resourceScoringFunction((2+6),8)
  15.                = (100 -((8-8)*100/8))
  16.                = 100
  17.                = rawScoringFunction(100)
  18.                = 10
  20. NodeScore   =  (5 * 5) + (7 * 1) + (10 * 3) / (5 + 1 + 3)
  21.             =  7

What’s next