These Technical Requirements (TR) define the ways in which you may or may not make use of the CNX services. This document will be regularly reviewed and revised in the light of operational experience to ensure maximum protection of service to CNX members.
last update December 2024
Physical Configuration
- No media converters are allowed, all connections must terminate to a transceiver module (SFP/SFP+/QSFP28) on a CNX switch.
- The connection should be a direct dark fiber connection from the CNX members router (or L3 switch) to the CNX switch, with no devices in between such as media converters or L2 switches.
Allowed Optical Interfaces for Connections
- Fiber Type: Only single-mode optical fiber (SMF) compliant with ITU-T G.652 standards is permitted for physical layer connections.
- Optical Transceiver Compatibility:
- The IXP supports connections exclusively using fiber transceiver (SFP/SFP+/QSFP28).
- Transceivers must operate on wavelengths compatible with the 1310 nm or 1550 nm spectrum, as per IEEE 802.3 specifications for Ethernet interfaces.
- Physical Interface Requirements:
- All connections must terminate using LC connectors for duplex operation.
- Fiber patch cords must be of sufficient quality to meet or exceed insertion loss and return loss requirements for the respective speeds (1GE, 10GE, 100GE).
- Prohibited Media:
- Multimode fiber (MMF) and any other optical interfaces not meeting the above specifications are not allowed.
- Copper-based Ethernet or any non-optical media is not supported.
MAC Layer
- Only specified ethertypes are allowed on the CNX VLANs. The policy is enforced with a VACL configured on the CNX switches.
- The following ethertypes are allowed:
- 0x0800: IPv4
- 0x0806: ARP
- 0x86dd: IPv6
- Frames with any other ethertypes are dropped on CNX switch ingress.
- All frames of a service forwarded from an individual CNX port shall have the same source MAC address.
- On the public peering network (VLAN 500) only 1 MAC address is permitted per port.
IP Layer
- Interfaces connected to public CNX peering ports shall only use IP addresses and netmasks (prefix lengths) assigned to them by CNX. In particular:
- IPv6 addresses (link & global scope) shall be explicitly configured and not auto-configured
- IPv6 site-local addresses shall not be used
- IPv6 router advertisements shall be disabled
- Standard IP MTU size = 1500
- For PNI ports, IP and MTU configuration is determined by the peer
Routing
- All exchange of routes across the open CNX peering network shall be via BGP4(+).
- AS numbers used in BGP4(+) sessions across the CNX peering network shall not be from range reserved for private use.
- IP address space assigned to CNX peering LAN shall not be advertised to other networks.
- All routes to be advertised in a peering session across CNX shall be registered in the APNIC or other public routing registry.
Route Server (RS)
- All peers in the public peering network must maintain a peering session with our route server.
- RPKI invalid routes will be dropped.
- All routes must have route object authorization (ROA) setup for the route to be propagated.
- All routes advertised shall be aggregated as far as possible, including across non-advertised address space.
- Additional information about our RS polices can be found at route server policy.
Forwarding
- Traffic shall only be forwarded to a CNX member when permission has been given by the receiving member either:
- by advertising a route across the CNX network (directly or via the routeserver)
- by explicit agreement between members.
- Traffic shall not be routinely exchanged on the public peering network between two CNX ports owned by the same CNX member. For site to site traffic please request a free virtual PNI.
CNX configuration
global mac access list
mac access-list extended ix-protocols permit any any 0x800 0x0 permit any any 0x806 0x0 permit any any 0x86DD 0x0
per port configuration
mac access-group ix-protocols in
default switchport configuration; if some configuration is missing, your port may not work or may be blocked
interface GigabitEthernet1/0/1 switchport nonegotiate no lldp transmit no lldp receive no cdp enable no keepalive spanning-tree portfast trunk spanning-tree bpdufilter enable spanning-tree bpduguard enable no shutdown