How Multipath load sharing aects route selection
During evaluation of multiple paths to select the best path to a given destination (for installment in the IP route table), the device performs
a nal comparison of the internal paths. The following events occur when load sharing is enabled or disabled:
• When load sharing is disabled, the device prefers the path with the lower device ID if the compare-routerid command is
enabled.
• When load sharing and BGP4 Multipath load sharing are enabled, the device balances the trac across multiple paths instead
of choosing just one path based on device ID.
Refer to How BGP4 selects a path for a route (BGP best path selection algorithm) on page 353 for a description of the BGP4 algorithm.
When you enable IP load sharing, the device can load-balance BGP4 or OSPF routes across up to four equal paths by default. You can
change the number load sharing paths to a value from 2 through 8.
How Multipath load sharing works
Multipath load sharing is performed in round-robin fashion and is based on the destination IP address only. The
rst time the device
receives a packet destined for a specic IP address, the device uses a round-robin algorithm to select the path that was not used for the
last newly learned destination IP address. Once the device associates a path with a particular destination IP address, the device will
always use that path as long as the device contains the destination IP address in its cache.
NOTE
The device does not perform source routing. The device is concerned only with the paths to the next-hop devices, not the entire
paths to the destination hosts.
A BGP4 destination can be learned from multiple BGP4 neighbors, leading to multiple BGP4 paths to reach the same destination. Each
of the paths may be reachable through multiple IGP paths (multiple OSPF or RIP paths). In this case, the software installs all the multiple
equal-cost paths in the BGP4 route table, up to the maximum number of BGP4 equal-cost paths allowed. The IP load sharing feature
then distributes trac across the equal-cost paths to the destination.
If an IGP path used by a BGP4 next-hop route path installed in the IP route table changes, then the BGP4 paths and IP paths are
adjusted accordingly. For example, if one of the OSPF paths to reach the BGP4 next hop goes down, the software removes this path
from the BGP4 route table and the IP route table. Similarly, if an additional OSPF path becomes available to reach the BGP4 next-hop
device for a particular destination, the software adds the additional path to the BGP4 route table and the IP route table.
Changing the maximum number of shared BGP4 paths
To change the maximum number of BGP4 shared paths, enter commands such as the following.
device(config)# router bgp
device(config-bgp-router)# maximum-paths 4
device(config-bgp-router)# write memory
Syntax: [no] maximum-paths num | use-load-sharing
The number parameter
species the maximum number of paths across which the device can balance trac to a given BGP4 destination.
The number value range is 2 through 8 and the default is 1.
When the use-load-sharing option is used in place of the number variable, the maximum IP ECMP path value is determined solely by
the value congured using theip load-sharing command.
Optional BGP4 conguration tasks
FastIron Ethernet Switch Layer 3 Routing
53-1003627-04 379
To Next Page
Loading...
Need help?
General
