cloudscaleKnowledge Base
Scale Servers
You can change the compute flavor of a cloud server at any time to adjust its performance to your current needs.
At cloudscale, you can choose from a wide range of compute flavors. A flavor specifies how much CPU and RAM resources are available to your server; for GPU servers, the flavor also determines how many GPUs are allocated to the server and how large the local NVMe scratch disk is.
To ensure that your servers can always deliver the right performance, even as your requirements change, you can "scale" them to a different (larger or smaller) flavor at any time.
To scale a server, it must be temporarily powered down. Then, in the cloud control panel, go to your server's detail view, select the desired new flavor on the "Compute" tab, and confirm by clicking the green "Scale" button at the bottom of the page.
At cloudscale, services are billed by the second. This means that your server will be billed at the price of the old flavor until the time of scaling, and at the price of the new flavor from that point on.
"Flex" and "Plus" flavors:
You can scale a server with a Flex or Plus flavor to any other Flex or Plus flavor. Select the new flavor on the "Shared vCPUs" and "Dedicated CPU Cores" tabs, and adjust the CPU-to-RAM ratio as needed using the "Change vCPUs" or "Change Cores" dropdown menus.
Once you confirm the new flavor, the scaling process takes only a few seconds; you can then power on the server again.
"GPU" flavors:
You can scale a server with a GPU (or GPUs) to any other GPU flavor with a scratch disk that is at least as large as the current one. On the "Dedicated GPUs" tab, select the desired RAM/GPU configuration, and, if necessary, adjust the number of CPU cores and the size of the scratch disk in the "Change Variant" dropdown.
When scaling, your GPU server may be migrated to a different physical host in the background. Since the contents of your local NVMe scratch disk must also be moved in the process, it may take some time before you can power on the server again.
Please also note that GPU resources are currently available to a limited extent, and scaling may fail due to a capacity bottleneck. If necessary, contact our support to review the options for your specific situation.