|In many applications, it is desired to build redundant links so that the failure of a radio will not take away the connectivity to a remote site. In a network where the transport mechanism is transparent bridging, you cannot just put two links up in parallel, because that creates a loop that will keep duplicating packets until it the unwanted packets fill up the entire network capacity. (This is called a packet storm.) This problem can be eliminated by installing a managed ethernet switch with the STP protocol, which selects a link to shut down to eliminate the loop.Ethernet switches come in several classes at different price points. The least expensive (typically under USD 100) are unmanaged switches. These are suitable for use in home networks, in very simple office networks and within a single room in a larger office building. These switches have no configuration at all. The next higher level (typically USD 100-200 for a 16-port switch) are called web-managed switches. They support 802.1Q VLANs and a small amount of bandwidth management. Some also support Spanning Tree Protocol (STP) to prevent loops in the network. At the time of writing this, these switches do NOT allow you to configure multicast filtering. Finally, fully managed switches tend to be in the price range of USD 300 and up. These switches have a very rich feature set and can be controlled by SNMP in addition to a serial port (command line) interface and a web interface. These switches all support several flavors of STP in addition to VLAN and SNMP. Because not all web-managed switches fully implement STP (amd RSTP) we recommend a fully managed switch (such as the D-Link 3028) for this function.
Example of a redundant network link:
STP has a few settable parameters. By setting the Hello timeout and the link hold time to their lowest value, this network will switch the link in 20 to 30 seconds after a failure.
|In this sketch, the B-D and C-E are two identical radio links between two sites.
B and C are at one site (“the central location”) and D and E are at the other site (“the remote location”).
G is a critical computer system at the remote location which must have assured connectivity to the central location even if one of the radios fail.
A is an unmanaged switch at the central location.
F is a managed swtich with STP at the remote location.
The two radio links are set to different RF channels, so that they do not interfere with each other, and they synchronize their cycles by means of heartbeat packet that pass through the unmanaged switch.
During normal operation, when both links are up, switch F detects that there is a loop between the ports connected to the radios D and E, and it shuts down one of these ports to eliminate the loop. If one of the radio links should fail, the loop goes away, and the switch will re-enable the disabled port to restore connectivity.