We are excited to announce another Kubernetes specific launch that is focused on enhancing visibility into the cost allocation of each cluster to Namespaces and Labels as a part of these infrastructure costs being assigned back to the business. Back in May we launched Idle Cost Distribution and Reporting for Containers in Core Analytics and last month we launched VM Family-Based Weightings for Container Cost Allocation. This approach improved how resources such as CPU and memory are weighted from a cost perspective,  better matching how the underlying Virtual Machines are charged by the cloud vendors.

With this launch users are now able to see the exact cost contribution of CPU, memory, GPU, storage, and network as calculated by Cloudability. This is surfaced through a set of new cost metrics that are selectable within the table of the Container Cost Allocation feature:

Figure 1: Selecting cost metrics within the Container Cost Allocation feature

For each resource type there are two metrics available. For example:

·      CPU Allocation – this is the dollar figure associated with the consumed CPU

·      CPU Fair Share – this is the dollar figure associated with CPU including the idle component

Using cost metrics at the cluster level

These metrics can be used at the cluster level to quantify the total cost of each resource type. For example, it is possible to quantify the CPU cost per cluster, both the consumed amount and total amount (including idle).

Figure 2: Reviewing memory and CPU cost per cluster

Note that if your clusters are not using GPU-backed VMs then the GPU cost metrics will be $0. If you choose to distribute the idle cost (option available towards the top of the UI) then adding up all the “Fair Share” cost metrics for a cluster will equal the “Total” metric, whereas the “Allocation” metrics sum to the “Utilized” metric. 

Using cost metrics at the Namespace or Label level

Probably the most valuable use case for these new cost metrics is to use them when enumerating Namespaces or Labels. This enables the user to know exactly why a specific Namespace or Label value was allocated a specific cost portion of a cluster. For example, if a given Namespace has been allocated $10,000 the user can add these cost metrics to the table and understand what part of that was driven by CPU consumption versus memory consumption etc.

Figure 3: Reviewing memory and CPU cost per Namespace for a specific cluster

Please note that these new cost metrics are available for historical data within the Container Cost Allocation feature.