Reference Architecture

This document lists the minimum reference architecture to get OpenStack installed with OpenDayLight. Wherever possible, additional resources will be stated.

Cloud Composition

The basic cloud will have 3 types of nodes:

  • Controller Node - Runs OpenStack services and the ODL controller.
  • Network Node - Runs the DHCP agent, the metadata agent, and the L3 agent (for SNAT).
  • Compute Node - VMs live here.

Usually each of the first 2 types of nodes will have a cluster of 3 nodes to support HA. It’s also possible to run the ODL controller on separate hardware than the OpenStack services, but this isn’t mandatory.

The last type of nodes can have as many nodes as scale requirements dictate.

Networking Requirements

There are several types of networks on the cloud, the most important for the reference architecture are:

  • Management Network - This is the network used to communicate between the different management components, i.e. Nova controller to Nova agent, Neutron to ODL, ODL to OVS, etc.
  • External Network - This network provides VMs with external connectivity (i.e. internet) usually via virtual routers.
  • Data Network - This is the network used to connect the VMs to each other and to network resources such as virtual routers.

The Control Nodes usually are only connected to the Management Network, unless they have an externally reachable IP on the External Network.

The other node types are connected to all the networks since ODL uses a distributed routing model so that each Compute Node hosts a “virtual router” responsible for connecting the VMs from that node to other networks (including the External Network).

This diagram illustrates how these nodes might be connected:

            Controller Node
           +-----------------+
           |                 |
 +-----------+192.168.0.251  |
 |         |                 |
 |         +-----------------+
 |
 |          Compute Node      +----------------+
 |         +---------------+  | Legend         |
 |         |               |  +----------------+
 +-----------+192.168.0.1  |  |                |
 |         |               |  | --- Management |
 | +~~~~~~~~~+10.0.0.1     |  |                |
 | |       |               |  | ~~~ Data       |
 | | +=======+br-int       |  |                |
 | | |     |               |  | === External   |
 | | |     +---------------+  |                |
 | | |                        +----------------+
 | | |      Network Node
 | | |     +-----------------+
 | | |     |                 |
 +-----------+192.168.0.100  |
   | |     |                 |
   +~~~~~~~~~+10.0.0.100     |
     |     |                 |
     |=======+br-int         |
     |     |                 |
     |     +-----------------+
+----+---+
|        |
| Router |
|        |
+--------+

Minimal Hardware Requirements

The rule of thumb is the bigger the better, more RAM and more cores will translate to a better environment. For a POC environment the following is necessary:

Management Node

CPU: 2 cores

Memory: 8 GB

Storage: 100 GB

Network: 1 * 1 Gbps NIC

Network Node

CPU: 2 cores

Memory: 2 GB

Storage: 50 GB

Network: 1 Gbps NIC (Management Network), 2 * 1+ Gbps NICs

Compute Node

CPU: 2+ cores

Memory: 8+ GB

Storage: 100 GB

Network: 1 Gbps NIC (Management Network), 2 * 1+ Gbps NICs