Dedicated host
A dedicated host represents virtualized resources that are running on dedicated hardware, exclusively for one customer. Customers having dedicated hosts do not need to share resources with other customers and gain better isolation and control.
A dedicated host can support either standard processing (with a certain degree of over-subscription) or dedicated processing (through CPU pinning). The customer selects the processing type for each dedicated host. The available configuration options for a dedicated host are shown in Figure 1.
Figure 1. Configuration options for dedicated hosts.
Table 1 gives an overview of the options in standard and dedicated processing.
Dedicated host with standard processing | Dedicated host with dedicated processing |
|---|
Standard processing relies on over-subscription/over-commitment of processing power. In practice, the processing power of that compute node is then shared between the customer’s instances. Suitable for workloads that do not require extreme CPU or networking (latency/pps) performance. Virtual machine flavors can be customized. Standard networking performance applies.
| Dedicated processing offers vCPUs that are exclusively pinned to a specific virtual machine. CPU arrangements for particular NUMA architectures are also supported. Suitable for workloads that require extreme CPU and/or networking (latency/pps) performance as well as better isolation. Virtual machine flavors can be customized. Accelerated networking performance is achieved through SmartNics and virtio-forwarder technologies. More details can be found in Accelerated networks.
|
Table 1. Processing options with a dedicated host.