Understanding how Capacity Units for SAP BTP, Kyma runtime are calculated
Recently, I was asked how the monthly sum of Capacity Units for SAP BTP, Kyma runtime is calculated. “Capacity Unit” is a metric which – in the case of Kyma runtime – combines two components into a monthly sum. Let me elaborate on each of them.
“Storage” is based on the persistent volume claims (PVC) Kyma customers make inside their runtime to store data in the cluster. These are 32GB steps of HDD storage which are used e.g. in this tutorial. The “Base Configuration” of Kyma claims 7 of those PVs which is covered in the monthly consumption of Capacity Units which you can’t avoid. These PVs are called
All other PVCs are charged on an hourly basis in steps of 32 GB. Each Persistent Volume adds 13 Capacity Units to your sum of 720 hours, i.e. 30 days.
With the migration of Kyma runtime to the open source version 2 of project “Kyma”, you can easily find all the PVCs at the root of the Kyma Dashboard in the menu “Storage” > “Persistent Volumes”.
These nodes are a split of the virtual machines you selected when setting up Kyma runtime. We want to provide you with most flexibility for VM sizes, hence the smallest common denominator is a node of 2 virtual CPUs and 8 GB of memory. All the possible virtual machines for SAP BTP, Kyma runtime are a multiple of this node size. E.g. the default 8vCPU/32GB RAM virtual machine is calculated as 4 of these smaller nodes.
The base setup of Kyma runtime requires two virtual machines, which calculates 2×4=8 nodes. Each additional virtual machine of 8 vCPUs and 32 GB RAM is adding 4 more nodes to the monthly sum.
In the new Kyma Dashboard, you can see right at the top of the entry page, how many VMs your Kyma runtime consists of. Unfortunately, in Kubernetes, a virtual machine is as well called “Node” which of course confuses with term “Node” in the Kyma estimator. But it’s not the same. A “Node” in the Kyma Dashboard is one virtual machine of the size you selected when setting up Kyma runtime.
More VMs active than you expect?
If you see more VMs listed in the Kyma Dashboard than you think there should be, ensure that your deployments don’t reserve too much CPU or Memory. Within the built-in Grafana open the dashboard “Kubernetes / Compute Resources / Cluster” for analyzing the actual CPU/Memory Utilization and the “Requests Commitment“.
With these two components – the number of additional persistent volumes > 7 and the number of virtual machines > 2 – you can use https://estimator-don4txf2p3.dispatcher.ap1.hana.ondemand.com/index.html to calculate the cost for 720 hours, i.e. 30 days.
Are you missing “Event Throughput”? This was just removed from the calculation. As we switched to an in-cluster eventing system, this is not charged anymore.