A common concept in high-level network architecture is the notion of several distinct planes of the network: forwarding plane, control plane, and management plane. The management plane of the network is made up of the systems required to operate, administer, monitor, and maintain the network. The nodes in a management plane may include monitoring servers, data collection servers, authorized administrator PC's, and servers hosting the control plane of the network being managed.
While many clouds and data center environments enjoy the luxury of a cleanly separate network forwarding plane to connect these management plane systems, this is not common other networks such as a WAN. In these environments the management of the network must be carried out within it's own network forwarding plane. This describes a practice for 128T networks to give the management plane first-class citizenship within the forwarding plane. It describes the management plane systems as a proper network tenant having requisite services with routing, inclusive of the Linux servers hosting the control plane for 128T router nodes.
Every endpoint on a 128T network is given tenancy, and the endpoints that make up management plane are no exception. When network management is carried out within the forwarding plane it should be given it's own proper tenant, in order to cleanly segment it from other endpoints. For example, the following would establish a tenant called
OAM for the authority. Endpoints anywhere in the network belonging to the management plane that will need to send sessions, such as administrator PC's, monitoring servers, and 128T node hosts, can be given
There are often many services required within the management plane of a network. This includes services for things like DNS, NTP, SSH/HTTPS element management, software distribution servers, IPFIX collectors, syslog receivers, and so on. Each one of these should be modeled with proper services providing access from the management tenant.
Because every 128T router node host may be the target of management sessions, each node must be given a service such as the following. For simple IP address management (IPAM) and routing, ideally contiguous blocks of addresses can be allocated for 128T router hosts. In this example, we'll assume that
10.128.0.0/16 is reserved in IPAM for 128T node hosts.
128T Router Nodes
The Linux kernel of every 128T router node in the network will be the host of inbound sessions to a management plane service, and will originate outbound sessions to other management plane services. For certain procedures, like inventory discovery, some management plane systems may use the source address of an inbound connection as a key to send outbound sessions back to the address. Because of this, sessions should be free of any source or destination NAT whenever possible.
The following configuration on a 128T router node would provide it with a host Kernel Network Interface (KNI) for attachment to the management plane. It is made the default route in the kernel route table, given management plane tenancy for outbound sessions, and set up with the Linux side of the KNI having the address corresponding to the management plane service for inbound sessions to the node. Review the Host Services configuration guide for more details.
In the 128T configuration of a host KNI, you are describing the 128T forwarding plane side of the KNI. The configuration of the address in the
gateway setting is automatically given to the Linux side of the KNI. In other words, the service address for inbound sessions to the node should be set to the
gateway in the address configuration for the 128T KNI.
In order to optimize address utilization, the pool of addresses which are allocated for use as 128T node management services, can be configured as KNI gateways with their
/31 complement as the 128T forwarding plane address. For example, the allocation of
10.128.0.3 to a set of nodes could have addresses configured like this:
- Node1 with a management service address
- Node2 with management service address
Throughout the topology of 128T routers, management services must be distributed using STEP and routed using Secure Vector Routing. This means that static service routes are created where the services need to egress, and dynamically routed from wherever they ingress.
For 128T node management services, this would mean that each router would be provisioned with a service route like the one below:
Where 128T routers are interconnected with other external routing domains, the service addresses for 128T node management may need to be advertised using another routing protocol such as BGP or OSPF. This ensures that they have a route for packets which are part of forward flows on sessions going to 128T node management services, and reverse flows for sessions originating from 128T host nodes.
Simple and effective management of a network is a key to deployment success. With this practice of management built into the forwarding plane of a 128T network, management services, including the 128T node hosts themselves, can be consistently accessed with all the benefits that 128T networking provides. Network administrators, operations personnel, monitoring systems, and 128T hosts themselves are given tenancy for secure access to the services they need.