The compute power (CPU power) available for cloud resources is stated in compute units (CUs). One CU corresponds to the physical processor power of 100 MHz.
When determining the scale, you should also consider which applications will have to be run on the vMs. In that regard, take into account all relevant information provided by the software producer. Swisscom recommends assigning no more vCPUs to any vM than the primary application operated on it will support. If too many vCPUs are assigned to a system, reduced performance for the system will be the result in most cases. The reason being that, equipped with several vCPUs, the hypervisor has to reserve more physical compute power (cores) at the same time. One large sector is processed less efficiently than many small parallel sectors. However, since the application cannot address all vCPUs, the reserved power is used only partially. At that point, additional vMs cannot use the currently unused, reserved portion of compute power neither.
Resource allocation for Dynamic Data Centers
As stated above, the cloud architecture supports two models of how the compute power is provided to virtual servers running in a Dynamic Data Center: the “Allocation Pool” model and the “Pay as you go” model. You can chose between the two when ordering a DDC.
NOTE: Once a DDC has been created, a change of the resource allocation model is no longer possible.
“Allocation Pool” Model
CUs are administered per DDC in a resource pool. The compute power available in a pool is dynamically allocated to the virtual systems based on their respective load behaviour. This mechanism ensures that a data center's pool resources can balance out the individual systems' peak loads. In order for a virtual system to be activated, at least three CUs per virtual CPU (vCPU) have to be available. A vCPU can apply a maximum of 24 CUs.
When determining the scale of a Dynamic Data Center's resource pool, you can rely on the following reference values to obtain an initial estimate:
|Processor power||Example||Number of CUs per vCPU|
|Low processor requirements||DHCP server||5|
|Medium processor requirements||Middleware server||10|
|High processor requirements||Multitenant application||15|
The information in the table refers to one DDC, in which virtual machines (vMs) are operated with a total of at least 10 vCPUs. In the case of a lower number of vCPUs, additional CUs will have to be provided to the pool because the peak loads of individual systems are more difficult to compensate for. Furthermore, performance depends on the workloads running in a DDC. Furthermore additional CUs are required if the workloads are volatile and resource reallocation needs to be adapted on a very frequent basis.
“Pay as you go” Model
The main difference to the “Allocation Pool” model is that CUs are assigned per vCPU and not shared in a resource pool. When ordering a DDC you can specify the performance class per vCPU. The following performance classes are available:
- 6 CUs/vCPU
- 12 CUs/vCPU
- 18 CUs/vCPU
- 24 CUs/vCPU
NOTE: Based on design considerations by VMware only one vCPU performance class can be used within the same DDC.
Resource allocation for Dynamic Servers
The CU pool size can only be configured with DDCs. With a DS, the number of vCPUs predefined for the applicable performance category is available for a virtual server (Swisscom is responsible for managing the underlying pool resources). The number of vCPUs per server depends on the server size selection, ranging from XXS to XXXL, and it varies between one and four vCPUs.
The number of assigned vCPUs with Dynamic Servers:
|Performance category||Compute power|
|XXS / XS / S||1 vCPU (1)|
|M / L / XL||2 vCPU (1)|
|XXL / XXXL||4 vCPU (1)|
(1) vCPU power: 1.6 GHz