Chapter 4. MidoNet gateway nodes

In addition to the hardware required to run OpenStack and MidoNet, you need one or more servers to run at the edge of the network and connect your OpenStack deployment to the external network. These servers are referred to as "Gateway Nodes".

Gateway Nodes will handle more traffic (to and from external networks) than Compute nodes so you must make sure that you allocate more memory to the Gateway Nodes hosts.

 Gateway node requirements

MidoNet interfaces directly with upstream routers via physical connections.

GW nodes need at least two physical network interfaces, one on the internal network, and the other connected to the upstream router. Depending on the use case and environment, the GW nodes may be handling much incoming and outgoing traffic. Therefore, we highly recommend NICs with multi-queue and MSI-X support, such as ones based on the 1Gbps Intel™ 82576 Ethernet controller, or the 10Gbps Intel 82599 Ethernet controller. Using such a NIC enables us to process receive events on multiple queues, and to handle IRQ events on multiple cores.

 Gateway node connectivity

The MidoNet cloud deployment needs to be assigned a global IP range, which will be used by clients to access the VMs from the Internet.

MidoNet can use static routing, but we generally prefer to use BGP dynamic routing for fast fail over. All GW nodes advertise the same IP range via BGP so the upstream network should balance incoming flows over them. Typically, users assign a private Autonomous System (AS) to the MidoNet deployment, and set up BGP sessions for each edge.

Questions? Discuss on Mailing Lists or Chat.
Found an error? Report a bug.