Cisco SD-WAN Cloud onRamp for Colocation solution is a flexible architecture that securely connects to enterprise applications that are hosted in the enterprise data center, public cloud, private or hybrid cloud to an enterprise’s endpoints such as, employees, devices, customers, or partners. This functionality is achieved by using Cisco Cloud Services Platform 5000 series (CSP- 5444, CSP-5456) as the base Network Function Virtualization (NFV) platform that securely connects enterprise's endpoints to applications. By deploying Cisco SD-WAN Cloud onRamp for Colocation solution in colocation centers, customers can virtualize network services and other applications, and consolidate them into a single platform.

The components of Cisco SD-WAN Cloud onRamp for Colocation solution are:

• CSP-5444, CSP-5456—Cisco CSP is an x86 Linux hardware platform that runs NFVIS software. It is used as the compute platform for hosting the virtual network functions in the Cisco SD-WAN Cloud onRamp for Colocation solution. The whole solution can scale horizontally. You can have up to eight Cisco CSP devices. Depending on the load requirement, you can have anywhere from two to eight compute platforms in a cluster.

• Cisco Network Function Virtualization Infrastructure Software (NFVIS)—The Cisco NFVIS software is used as the base virtualization infrastructure software running on the x86 compute platform.

• Virtual Network Functions (VNFs)—The Cisco SD-WAN Cloud onRamp for Colocation solution supports both Cisco-developed and third-party virtual network functions.

• Physical Network Functions—A Physical Network Function (PNF) is a physical device that is dedicated to provide a specific network function as part of a colocation service chain such as router, firewall.

• Network Fabric—Forwards traffic between the VNFs in a service chain by using a L2 and VLAN-based lookup.

• Mangement Network—A separate management network connects the NFVIS software running on the Cisco CSP systems, the virtual network functions, and the switches in the fabric. This management network is also used for transferring files and images into and out of the systems. The Out of Band (OOB) management switch configures the management network.

• VNF Network Connectivity—A VNF can be connected to the physical network by using either Single Root IO Virtualization (SR-IOV) or through a software virtual switch. A VNF can have one or more virtual network interfaces (vNICs), which can be directly or indirectly connected to the physical network interfaces. A physical network interface can be connected to a software virtual switch and one or more VNFs can share the virtual switch. The Cisco SD-WAN Cloud onRamp for Colocation solution manages the creation of virtual switch instances and the virtual NIC membership to create connectivity.

• Physical Network Function Network Connectivity—A PNF can be connected to the Cisco Catalyst 9500-40X or Cisco Catalyst 9500-48Y4C switch port, which is kept free towards backend.

• Service Chains—In Cisco SD-WAN Cloud onRamp for Colocation solution deployment, the traffic between the VNFs (with SR-IOV) running on the Cisco CSP systems is connected by service chaining externally through Catalyst 9500 switches.

• Cisco Colo Manager (CCM)—This component is a software stack that manages a colocation. In this solution, CCM is hosted on NFVIS software in a docker container. A single CCM instance per cluster is brought up in one of the Cisco CSPs after activating a cluster.

• Orchestration using Cisco vManage— Cisco vManage is used for orchestrating the Cisco SD-WAN Cloud onRamp for Colocation solution.

Essentially, you can purchase the CSP and Catalyst 9500 devices, add them in colocation centers, power them, and cable them. These devices automatically boot up and are managed by Cisco vManage. Next, design service chains, build security policies and application policies to impact the network traffic.

Deactivate cluster workflow is not supported: The workflow to deactivate a cluster and then subsequently activate the cluster is not supported through vManage. Therefore, it is recommended to delete and recreate the cluster, instead.