To configure your multi-site topology, you should understand the recommended topologies and the topologies to avoid.
A fully connected mesh network topology is one in which all sites know about each other. This is a robust configuration, as any one of the sites can go down without disrupting communication between the other sites. A fully connected mesh topology also guarantees that no site receives multiple copies of the same message.
A fully connected mesh with three sites is shown in this figure. In this scenario, if site 1 sends an update to site 2, site 2 forwards to site 3. If site 1 sends an update to sites 2 and 3, neither forwards to the other. This is likewise true for any other initiating site. If any site is removed, the remaining two are still fully connected.
A ring topology is one in which each site forwards information to one other site, and the sites are connected in a circular manner. This figure shows a ring with three sites. In this topology, if site 1 sends updates to site 2, site 2 forwards the updates to site 3. No updates are forwarded to the original sender, so site 3 does not send the updates back to site 1.
A ring topology guarantees that every site receives one copy of each message sent by any site. In a ring, every site must stay up to maintain the connection. The failure of any site breaks the ability for updates to reach all sites. If site 2 went down, for example, site 3 could send to site 1, but site 1 could not send to site 3.
There are numerous hybrid network topologies. Some of the sites are fully connected, while others form a ring.
The following figure shows a hybrid topology that forms a ring, with an extra connection that fully connects sites 1 and 3.
With this hybrid topology, if site 2 went down, it would not affect communication between sites 1 and 3. If site 3 went down, however, site 2 would not be able to send to site 1.
A second example hybrid topology is shown in the figure below. In this tree topology with site 1 as the root of the tree, sites 2 and 3 do not communicate with each another. This topology works for an application in which site 1 is a producer and the consumers (sites 2 and 3) have nothing to gain from being connected to each other. This topology also guarantees that no site receives the same update twice.
Topologies in which the same update may be delivered twice to a particular site do not work and are unsupported.
The DAG topology shown in this figure is an example of an unsupported technology. Site 4 will receive more than one copy of the same message when site 1 sends a message to sites 2 and 3, and sites 2 and 3 each forward the message to site 4.