# Workloads

The Workloads tab gives an overview of every workload running in your cluster. You can see what images are being used, resource requests and limits, and cost estimates.

workloads

# Cost Estimates

When you first load the workloads page, you'll be prompted to provide some information about your cloud provider and instance type settings. We'll use this information to inform our cost estimates.

workloads

Currently, we only support setting a single instance type. If you're using multiple instance types in your cluster, you can simply pick the most representative type.

Once we know your node size and cost per node, we assume half the cost is allocated to memory, and half the cost is allocated to CPU. We then extrapolate to determine cost-per-GB-RAM and cost-per-CPU.

To determine the cost of a particular workload, we offer two strategies:

  • conservative - this takes into account the potential waste incurred by memory- or CPU- intensive workloads, if Kubernetes is unable to bin-pack efficiently. It is calculated as 2 * max(cpu_cost, memory_cost)
  • optimistic - this assumes Kubernetes can bin-pack your workloads efficiently. It is calculated as cpu_cost + memory_cost

If you have spent time optimizing your node size, or if you're running a large variety of workloads that are small relative to your node size, the optimistic strategy will probably be more accurate. Otherwise, we recommend the conservative strategy.

You can read more about cost estimation on our blog