Skip to main content

Select VM size, pool and GPU

Choosing the right ccloud³ VM plan depends on your workload. An oversized ccloud³ VM would not make full use of its resources and would cost more, whilst an undersized ccloud³ VM could suffer from reduced performance or errors when CPU or memory utilisation is at full capacity.

To help you choose the best ccloud³ VM plan for your use case, this article explains the differences between shared and dedicated CPUs, provides a detailed overview of each ccloud³ VM plan, and concludes with recommendations to help you make a data-driven decision.

You can also upgrade a ccloud³ VM to a larger plan after it has been created, including upgrading to a larger ccloud³ VM plan of a different type. For example, you can upgrade from a Basic ccloud³ VM plan to a larger, CPU-optimised ccloud³ VM plan. The price list for ccloud³ VMs can be found on the pricing overview page.

Shared CPU vs. Dedicated CPU​

A ccloud³ VM is a virtual machine (VM) that is allocated resources such as CPU, RAM and hard disk space by a physical host.

A hypervisor, also known as a virtual machine monitor, ensures that the ccloud³ VMs running on a physical host each receive their own virtual resources, such as vCPUs.

A vCPU is a unit of processing power that corresponds to a single hyper-thread on a processor core. A modern, multi-core processor has several vCPUs.

The ccloud³ VM plan you have selected determines the amount of resources allocated to the ccloud³ VM. RAM and local SSD storage are permanently allocated to your VM. CPU performance varies depending on the pool: the Power pool offers a guaranteed minimum clock speed of 3.5 GHz, whilst the Worker pool operates at 2.7–3.4 GHz.

Dedicated CPU-ccloud³ VMs have guaranteed access to the full hyper-thread at all times. With shared CPU-ccloud³ VMs, the hyper-thread allocated to the ccloud³ VM may be shared amongst several other ccloud³ VMs. If a shared CPU-ccloud³ VM experiences a higher load, the hypervisor dynamically allocates additional hyper-thread(s).

However, the number of CPU cycles available to the hypervisor depends on the workload of the other ccloud³ VMs sharing this host. If these neighbouring ccloud³ VMs are under heavy load, a ccloud³ VM might be allocated fractions of hyper-threads rather than dedicated access to the underlying physical processors. In practice, this means that shared CPU-ccloud³ VMs may have access to full hyper-threads, but this is not guaranteed.

Data-driven decision-making​

Before you decide on a specific ccloud³ VM type, we recommend that you benchmark your workload and carry out load testing to see how it performs under simulated load. For bursty applications or batch jobs, you should monitor resource utilisation when the load is at its expected peak, particularly on shared CPU Basic ccloud³ VMs. If you find that your app’s performance is too variable for your production requirements, you should consider a ccloud³ VM type with dedicated vCPUs.

GPU​

With ccloud³ VMs, you can choose between the following GPUs

  • NVIDIA RTX A4000 (16 GB VRAM): The all-rounder for CAD, 3D modelling, video editing and AI-powered image processing

  • NVIDIA Quadro RTX 6000 (24 GB VRAM): For 3D animation, VFX, VR and simulation

  • NVIDIA A100 (40 GB or 80 GB VRAM): For AI model training, HPC, and big data and analytics

  • NVIDIA RTX 6000 Ada (48 GB VRAM): For real-time ray tracing, deep learning and research

CPU and RAM​

With ccloud³ VM graphs, you can obtain more information about the CPU load and memory usage of your ccloud³ VM:

  • If CPU utilisation is high and memory utilisation is also significant, you should scale both vCPUs and memory and use a balanced General Purpose ccloud³ VM plan.

  • If your CPU utilisation is high but your memory usage is very low, you could save money with a CPU-Optimised ccloud³ VM plan.

  • If memory usage is high (possibly at maximum capacity and being swapped to the hard disk) but CPU usage is low or moderate, you should scale the memory and use a Memory-Optimised ccloud³ VM plan.

  • If CPU or memory utilisation is low to moderate most of the time, but occasionally spikes and reaches resource limits, you should consider shared CPU Basic ccloud³ VMs and scale the limiting resource accordingly.

Hard disk​

All ccloud³ VMs from centron include varying amounts of local solid-state drive (SSD) storage. If you require additional storage, you can use network-attached block storage to attach additional volumes to a ccloud³ VM, or use object storage to offload files and associated metadata.

Network​

External data traffic is charged on a usage basis per GB – both incoming and outgoing traffic (see traffic calculation). There is no free traffic allowance based on VM size. Internal traffic between centron resources is free of charge.

Monitoring allows you to track both hard disk and bandwidth usage, in much the same way as you would monitor CPU and memory usage.

Last updated on