Hardware guidelines

Size your Cloud Management Platform (CMP) deployment based on the workloads you expect it to manage

2 minute read

Processing requirements depend on active workloads. Memory requirements depend on the total size of transactions and the objects being managed. The application itself has minimal storage requirements, but the underlying storage backend should have a relatively high-performance disk subsystem.

Deployment shapes

The two shapes below correspond to the two designs described in System architecture.

Single node is appropriate for most initial production deployments and for development and testing environments. It is one appliance running all CMP roles.
Distributed (HA) is for production environments with consistent, high workloads. This could be a large number of transactions, blueprints, managed VMs and objects, or a combination of the three. It splits the roles across three web servers and three job engines.

The CMP installer supports only the x86_64 architecture.

Hardware specifications

These are guidelines, not limits

The values below are neither hard limits nor maximum values. They give an idea of how to size your environment based on your expected managed servers, concurrent builds, and number of blueprints. Many factors contribute to performance and capacity, including the age and sizing of the underlying virtualization hardware and storage, as well as environmental factors such as security, firewall measures, and network speed and integrity.

Deployment	CPU	Memory (GB)	Disk (GB)	AWS	Azure	Users	Groups	Managed servers/objects	Concurrent builds (UI)	Concurrent builds (API)	Blueprint capacity (UI)
Single node	4	16	40	`m5.xlarge`	`Standard_D4as_v4`	7500	500	40,000	250	500	50
Distributed (HA)	8	32	40	`m5.2xlarge`	`Standard_D8as_v4`	10,000	800	60,000	500	1000	75

How to read this table

CPU, memory, and disk are per node. In the distributed deployment, each node is sized this way.
AWS and Azure instance types are examples that meet the CPU and memory figures, not the instance types you must use.
Managed servers/objects is the maximum tested amount with a 24-hour VM sync cycle. You can exceed it by increasing the sync interval. Adding job engine servers also increases capacity.
Concurrent builds refers to a single order containing that number of VMs.

Deployment composition

Single node: 1 appliance running all roles
Distributed (HA): 3 web servers and 3 job engines

These figures describe capacity, not entitlement. The number of servers you are licensed to manage is set by your license. See Product license.

Set up high availability: Configure the nodes that make up a distributed deployment.
Network connectivity: Ports to open between the nodes and the services they depend on.
Product license: How many servers your license entitles you to manage.