Decomposing the compute power

The compute service in Nova is considered to be the core component of OpenStack. Understanding how to scale out the workload among several identical compute nodes might require to briefly revisit the building blocks of Nova:

• nova-compute: This runs on the compute node as described in Chapter 1, Inflating the OpenStack Setup. It is responsible for communicating with the hypervisors. Nova-compute interacts with each hypervisor by means of drivers. It creates compute resources by picking up requests from the message queue.
• nova-scheduler: This runs on the cloud controller as described in Chapter 1, Inflating the OpenStack Setup. It is responsible for finding the right placement (physical server) of the initiated request to create a VM. The request will be left in the message queue along with additional information regarding the server information where the nova-compute service will create the compute resource.
• nova-api: This runs on the cloud controller as described in Chapter 1, Inflating the OpenStack Setup. It is responsible for handling API calls from other services through the messaging queue service.
• nova-conductor: This runs on the cloud controller as described in Chapter 1, Inflating the OpenStack Setup. It is responsible for managing access to the database for read/write operations for security and data coherence reasons.
• nova-consoleauth: This runs on the cloud controller as described in Chapter 1, Inflating the OpenStack Setup. It is responsible for providing authentication to the VNC console by the means of the VNC protocol.
• Metadata service: Optionally, this can run on the compute node as described in Chapter 1, Inflating the OpenStack Setup. It is responsible for booting a virtual machine with a custom configuration that will be consumed by the compute service.

The following diagram illustrates an overview of the different components forming the OpenStack compute service: