Understanding BGP as a Service
The BGP as a Service (BGPaaS) feature allows a guest virtual machine (VM) to place routes in its own virtual routing and forwarding (VRF) instance using BGP.
Contrail BGPaaS Features
Using BGPaaS with Contrail Networking requires the guest VM to have connectivity to the control node and to be able to advertise routes into the VRF instance.
With the BGPaaS feature:
The vRouter agent is able to accept BGP connections from the VMs and proxy them to the control node.
The vRouter agent always selects one of the control nodes that it is using as an XMPP server.
Contrail Networking provides route export functionality for BGPaaS sessions. The next hop for all routes advertised to the tenant VM is set to the default gateway address of the subnet of the tenant VM. This allows the tenant BGP implementation to be relatively simple, by not requiring support for recursive resolution of BGP next hops.
The BGPaaS object is associated with a virtual machine interface (VMI), not just a virtual machine (VM), which enables a tenant VM to have BGP sessions in multiple virtual networks, if required.
BGPaaS in Contrail Networking has the following features:
By default, all BGPaaS sessions are configured to have bidirectional exchange of routes. The Boolean property bgpaas-suppress-route-advertisement ensures no advertisement of routes to the tenant VM.
If inet6 routes are being advertised to the tenant VM, they are advertised with the IPv6 subnet's default gateway address as the BGP next hop. A Boolean property, bgpaas-ipv4-mapped-ipv6-nexthop, causes the IPv4 subnet's default gateway, in IPv4-mapped IPv6 format, to be used instead as the next hop.
If multiple tenant VMs in the same virtual network have BGPaaS sessions and they use eBGP, the standard BGP AS path loop prevention rules prevent routes advertised by one tenant VM from being advertised to the other tenant VMs. The as-override field, added to the existing
BgpSessionAttributesin the BGPaaS object, causes the control node to replace the AS number of the tenant VM with it's own AS number, when advertising routes learned from a tenant VM to another tenant VM in the same virtual network. The tenant VM does not need to implement any new functionality.
Contrail Networking also provides support for high availability (HA) architectures, BGPaaS supports control node zone selection, with options available to configure BGPaaS control node zone peers.
This capability enables you to set up primary and secondary control node zones, which can have one or more control nodes. The reason for this is because BGPaaS is often being relied upon to provide routing to and from VNFs, which are comprised of several nodes across different computes, and the VNFs usually rely upon two BGP peers for HA. These control node zone features increase the robustness and failover capabilities for BGPaaS in Contrail.
The following are caveats:
BGP sessions must use IPv4 transport.
The VNF must support RFC 2545, Use of BGP-4 Multiprotocol Extensions for IPv6 Inter-Domain Routing, to carry IPv6 routes over the IPv4 peer.
Only IPv4 (inet) and IPv6 (inet6) address families are supported.
The following features and properties are supported in Contrail Networking for BGPaaS configuration:
Global-System-Config has an option to add, modify, or delete control node zones
Control-Node-Zone has an option to add, modify, or delete control nodes
Control node has an option to add, modify or delete a control node zone and it can have only one control node zone
BGPaaS has an option to add, modify, or delete a primary or secondary control node zone
If control node zone has more than one control-node, selection of control-node for BGP Peering is random in a control node zone
Using just one control node in each zone, VNF can predictably establish bgp-peering to that particular control node.
BGPaaS Use Cases
This section provides example scenarios for implementing BGPaaS with Contrail.
Dynamic Tunnel Insertion Within a Tenant Overlay
Various applications need to insert dynamic tunnels into virtual networks. Virtual network functions (VNFs) provide the function of tunnel termination. Tunnel termination types vary across application types, such as business VPN, mobility small site backhaul, VPC, and the like. The key requirement is that tunnels need to insert dynamically new network reachability information into the virtual network. The predominant methods of tunnel network reachability insertion use BGP.
BGPaaS allows the migration of brownfield VNFs into Contrail, preserving the application behavior and requirement for BGP, without rewriting the application.
Figure 1 shows the need to insert a dynamic tunnel into a virtual network.
Dynamic Network Reachability of Applications
The Domain Name System (DNS) is a widespread application that uses BGP as a mechanism to tune reachability of its services, based on metrics such as load, maintenance, availability, and the like. As DNS services are migrated to environments using overlays, a mechanism to preserve the existing application behavior and requirements is needed, including the ability to announce and withdraw reachability to the available application.
This requirement is not limited to DNS. Other applications, such as virtualized evolved packet core (vEPC) and others, use BGP as a mechanism for network reachability based on availability and load.
Liveness Detection for High Availability
Various keepalive mechanisms for tenant reachability have been provided by network components such as BGP, OSPF, PING, VRRP, BFD, or application-specific mechanisms. With BGP on the vRouter agent, BGP can be used to provide a liveness detection mechanism between the tenant on the local compute node and the services that the specific tenant VM is providing.