数据中心

面向互联网交换点(IXP)的 OcNOS 数据中心:设计与配置 Guide

An Internet Exchange Point (IXP) is a shared network facility where multiple networks — ISPs, CDNs, cloud providers, and enterprises — interconnect to exchange traffic directly rather than routing through transit providers. The network infrastructure of an IXP has specific design requirements that differ from both enterprise and service provider networks.

IXP Network Design Requirements

  • Layer 2 fabric reliability — most IXPs use a shared L2 Ethernet fabric where all participants connect; the fabric must be loop-free, highly available, and support hundreds to thousands of participant VLANs
  • BGP route table scale — route server infrastructure processes the full internet routing table (1M+ prefixes) from hundreds of participants simultaneously
  • Port density — large IXPs need hundreds of 1G, 10G, and 100G ports; dense platforms that minimize rack space and power are essential
  • Traffic isolation — participant traffic must be strictly isolated; cross-participant L2 leakage is a security incident
  • Operations simplicity — IXPs are often operated by small teams managing large infrastructure; automation and simple operational tooling are critical

OcNOS-DC for IXP Architecture

IXP architecture: route servers connected to a Layer 2 VLAN fabric. Participant routers from ISP-A, ISP-B, CDN-C connect via 10G and 100G ports. OcNOS switches form the peering LAN fabric with RSTP and per-VLAN isolation.
IXP architecture on OcNOS-DC: two redundant fabric switches form the shared peering LAN. Route servers peer with all participants via BGP. Each participant is isolated to its VLAN; private VLANs enable bilateral peering between specific pairs.

OcNOS-DC IXP Configuration

! OcNOS-DC -- IXP peering LAN fabric configuration
!
configure terminal
!
! Step 1: create the peering VLAN (shared by all participants)
vlan database
  vlan 100 bridge 1 state enable
  exit
!
! Step 2: configure the participant-facing access ports
interface xe1
  description ISP-A (AS64501)
  switchport
  bridge-group 1
  switchport mode access
  switchport access vlan 100
  spanning-tree portfast            ! immediate forwarding on link-up
  storm-control broadcast level 1   ! limit broadcast to 1% of bandwidth
  exit
!
! Step 3: SVI for route server connectivity
interface vlan1.100
  ip address 193.0.0.1/24           ! IXP peering LAN address
  no shutdown
  exit
!
! Step 4: BGP route server (runs separately, not on the fabric switch).
! The route server peers with every participant on the peering LAN.
router bgp 65000
  bgp router-id 193.0.0.1
  neighbor 193.0.0.10 remote-as 64501   ! ISP-A
  neighbor 193.0.0.11 remote-as 64502   ! ISP-B
  neighbor 193.0.0.20 remote-as 64503   ! CDN provider
  exit
!
commit
exit
!
! Verification:
show mac address-table bridge 1 vlan 100
show bgp summary
show spanning-tree

IP Infusion Engineering Team

分享