Node Group Capacity View
Node Group Capacity View
#380
The Node Group Capacity page provides details of available capacity in your node groups, highlighting resource shortfalls that need to be addressed. You can also see details about waste in your environment, including information about the primary constraint and the optimal instance type that could be used when the environment is optimized.
Figure: Locating the Node Group Capacity View
You can create filters at the column or table level. See Using Filters. The capacity report provides the following details:
Table: Node Group Capacity Report
|
Column Name |
Description |
|---|---|
|
Cluster |
The cluster name containing the node group. By default, this table is sorted by Cluster and then Node Groups, in alphabetical order. |
|
Node Group |
The node group name. For nodes that do not belong to a node group, node group value will be <cluster-name>-default”. |
|
Primary Node Type |
The most common instance type. For example, if there are 20 r6i.large instances and 10 m6i.large instances, the primary node type is r6i.large. In cases where the instance type cannot be determined (i.e. bare-metal, private cloud environments) the primary node type is indicated as: CPU x memory size in GB, e.g. 8x16. A dash (-) indicates that node instance type could not be determined. |
|
Unique Nodes |
The total number of unique nodes that are part of this node group, during the past 7 days of history. The node ID is used to determine if a node is unique. |
|
Average No. of Nodes |
The average number of nodes in the node group, during the past 7 days of history for in-service instances with non-zero workload. For clusters (i.e. <cluster-name>-default) that are composed of nodes that are not part of a node group, the average number of nodes is calculated as: Total Node Hours/Total Hours in time period (7 days). For example, if there are 20 nodes, that are not part of any node group, and in total there are 200 node hours of utilization, then: 200 node hours/168 hours in 7 days = 1.19 Avg number of nodes. |
|
No. of Nodes at Max Pod Capacity |
The number of nodes in the node group where the peak number of pods that can be run is equal to the number of allocatable pods during the past 7 days of history |
|
This is the average number of cores available for your container workloads. This value is determined from the number of hours each node was running in the last 7 days and using that value to calculate how many cores are available on average. |
|
|
CPU Request |
The percentage of allocatable CPU capacity that is allocated to CPU requests. |
|
Peak CPU Utilization (%) |
The peak CPU utilization of the busiest node in the group. |
|
Average CPU Utilization (%) |
The average CPU utilization, during the past 7 days of history. This value is weighted by the number of hours that each node in the group ran. The CPU utilization is multiplied by the number of running hours for the node before calculating the average. |
|
This is the average amount of memory available for your container workloads. This value is determined from the number of hours each node was running in the last 7 days and using that value to calculate how much memory is available on average. |
|
|
Memory Request (% Capacity) |
The percentage of allocatable memory capacity that is allocated to memory requests. |
|
Peak Memory Utilization (%) |
The peak memory utilization of the busiest node in the node group |
|
Average memory Utilization (%) |
The average memory utilization during the past 7 days of history. This value is weighted by the number of hours that each node in the group ran. The memory utilization is multiplied by the number of running hours for the node before calculating the average. |
|
Primary Constraint |
Indicates the reason that more containers cannot be added to this node group. The following values are evaluated and whichever has the largest value is identified as the primary constraint:
|
|
Primary Node Type CPU (Cores) |
If a primary node type has been defined, then the number of cores is retrieved from the cloud model. Other nodes in node group may have different amounts of memory. If a primary node type has not been defined, use capacity_cpu. |
|
Primary Node Type Memory(GB) |
If a primary node type has been defined, then the number of CPU cores is retrieved from the cloud model. Other nodes in node group may have different numbers of CPUs. If a primary node type is not been defined, use capacity_mem. |
|
Fully Optimized Node Family |
Use this value to determine if your nodes are utilized as efficiently as possible. Both CPU and memory utilization are used to determine this ratio. Additionally, the following factors are taken into consideration when calculating the value.
If the node type is not one “Compute Optimized", “General Purpose”, or “Memory Optimized” then this value cannot be determined and is indicated as a dash (-). |
|
Surplus CPUs (Cores) |
The number of surplus cores currently provisioned. The surplus number of CPU cores for the node group, based on how many CPUs are utilized as compared to the total allocatable number of CPUs. If the node type is not one “Compute Optimized", “General Purpose”, or “Memory Optimized” then this value cannot be determined and is indicated as a dash (-). |
|
Surplus Memory (GB) |
The amount of surplus memory currently provisioned. The surplus memory of the node group based on how much memory is utilized as compared to the total allocatable amount of memory. If the node type is not one “Compute Optimized", “General Purpose”, or “Memory Optimized” then this value cannot be determined and is indicated as a dash (-). |
|
Estimated Waste ($/Month) |
This is the estimated wasted spend per month. If the value is negative your infrastructure may be under-provisioned. Calculated as: Fully Optimized Estimated Cost - Current Estimated Cost If the node type is not one “Compute Optimized", “General Purpose”, or “Memory Optimized” then this value cannot be determined and is indicated as a dash (-). |
|
No. of Container Manifests |
The number of container manifests in the node group being analyzed during last analysis run. Only containers that were part of the selected node group AND not part of another node group, in the past 7 days, are counted. |
|
Total |
The totals at the bottom of the table indicate the total for each column, for all pages, if you have a multi-page report. Vertical scroll bars may be displayed to scroll through the entire list, so lower rows may be momentarily hidden behind the row of Totals. |
Please observe the following for this set of data:
- Missing data is indicated with a dash (-);
- For bare metal cases, 4x16 is indicated as the primary node type. The current cost is determined using $20/core, $5/GB. The cost per core and per GB of memory is configurable. Contact [email protected] for details.
- g4dn.xlarge node group will always show '-' for the fully optimized and surplus columns because it is GPU family and not memory optimized, general purpose or compute optimized. Densify cannot accurately say what is happening there without GPU utilization data.
See the following topics for more details: