Load Balancing
 |
| |
When
two equal cost paths exist for a destination, an IP unicast routing protocol,
such as OSPF, will load-balance unicast traffic over the two links.
Load-balancing, without additional configuration, is not possible with multicast
routing protocols. The reason that load-balancing does not occur for multicast
traffic over equal cost links is because of the selection of the RPF interface.
Only one RPF interface can be selected for a multicast source and therefore all
multicast traffic must flow over that link. Multicast traffic flowing on the
other link will be rejected because it does not arrive on the RPF interface (see
Figure 9-6). |
| |
 |
| |
Figure 9-6: Multicast traffic is only accepted on one
link. |
|
 | |
 |
| |
In
order to achieve multicast load-balancing, we need to configure a tunnel between
routers A and B in Figure 9-6. All multicast traffic will flow across the tunnel
and the unicast routing protocols will load-balance across the actual physical
links (see Figure 9-7). Load-balancing occurs because we are encapsulating the
multicast traffic in unicast IP packets. Multicasting needs to be disabled on
the physical interfaces and enabled on the tunnel interface. |
| |
 |
| |
Figure 9-7: Load-balancing multicast traffic using a
tunnel. |
|
 | |
 |
| |
The
configurations for routers A and B are listed below: |
| |
 |
| |
Router A |
| |
| |
ip address 172.16.2.1 255.255.255.0 |
| |
| |
ip address 172.16.1.1 255.255.255.252 |
| |
| |
ip address 172.16.1.5 255.255.255.252 |
| |
| |
ip pim dense-mode (or sparse or sparse-dense
mode) |
| |
| |
tunnel destination 172.16.3.1 |
| |
 |
| |
Router B |
| |
| |
ip address 172.16.3.1 255.255.255.0 |
| |
| |
ip address 172.16.1.2 255.255.255.252 |
| |
| |
ip address 172.16.1.6 255.255.255.252 |
| |
| |
ip pim dense-mode (or sparse or sparse-dense
mode) |
| |
| |
tunnel destination 172.16.2.1 |
| |
 |
| |
Load-balancing will now occur over the two serial links, but the
mechanisms will be different, depending on whether the routers are
process-switching or fast-switching. For process-switching, the load-balancing
occurs with each packet using a round-robin method. Also, the packet counts on
each link will be the same. For fast-switching, load-balancing occurs with each
multicast flow because an (S,G) flow will be assigned to one of the physical
interfaces. | |
303 times read
|
Related news
|
| No matching news for this article |
|
Did you enjoy this article?
(total 0 votes)
|