Concept Guide

NOTE: An ecmp-group index is generated automatically for each unique ecmp-group when you configure multipath routes
to the same network. The system can generate a maximum of 512 unique ecmp-groups. The ecmp-group indices are
generated in even numbers (0, 2, 4, 6... 1022) and are for information only.
You can configure ecmp-group with id 2 for link bundle monitoring. This ecmp-group is different from the ecmp-group index
2 that is created by configuring routes and is automatically generated. These two ecmp-groups are not related in any way.
DellEMC(conf-ecmp-group-5)#show config
!
ecmp-group 5
interface twentyFiveGigE 1/2
interface twentyFiveGigE 1/3
link-bundle-monitor enable
DellEMC(conf-ecmp-group-5)#
Support for /128 IPv6 and /32 IPv4 Prefixes in Layer 3 Host Table
and LPM Table
IPv6 enhancements utilize the capability on platform to program /128 IPv6 prefixes in LPM table and /32 IPv4 prefixes in Host
table. Also host table provides ECMP support for destination prefixes in the hardware. The platform uses the hardware chip that
supports this behavior and hence they can make use of this capability.
CLI commands are introduced to move /128 IPv6 prefix route entries and /32 IPv4 prefix route entries from Host table to LPM
table and vice versa. When moving the destination prefixes from Route to Host table, there is a possibility of getting into hash
collision because the Host table on the device is a Hash table. In this scenario, a workaround does not exist for the user having
route entries programmed in host table.
When the command is issued, you are prompted with a warning message stating that the command configuration can take
effect on existing prefixes only when “clear ip route *” command is used. When you use the clear command, all the existing /32
IPv4 prefix route entries are reprogrammed in appropriate table. Also, all the other existing IPv4 entries are removed and
reprogrammed as a result of the clear command.
Dell EMC Networking OS releases earlier than Release 9.3(0.1) stores IPv6 /128 entries in Host table since it cannot be written
in LPM table, and IPv4 0/32 route entries are written in LPM table itself to support the ECMP since ECMP was not supported
in Host table. On the system, unified forwarding table (UFT) is enabled, and the host table size is bigger compared to the LPM.
When you move the IPv4 /32 route prefix entry in host table, more space is obtained that can be utilized for other route prefix
entries.
Support for ECMP in host table
ECMP support in the L3 host table is available on the system. IPv6 /128 prefix route entries and IPv4 /32 prefix entries which
are moved to host table can have ECMP. For other platforms, only the IPv6 /128 prefix route entries is stored in the L3 host
table without ECMP support.
The software supports a command to program IPv6 /128 route prefixes in the host table.
The output of show IPv6 cam command has been enhanced to include the ECMP field in the Neighbor table of Ipv6 CAM. The
sample output is displayed as follows, which is similar to the prefix table.
The following is the portion of the example output:
Neighbor Mac-Addr Port Vid EC
------------------------------------- -------------------------
[ 132] 20::1 00:00:20:d5:ec:a0 Fo 1/49/1 0 1
[ 132] 20::1 00:00:20:d5:ec:a1 Fo 1/50/1 0 1
Support for moving /128 IPv6 Prefixes and /32 IPv4 Prefixes
The software supports a command to program IPv6 /128 route prefixes in the route table. You can define IPv6 /128 route
prefixes in the route table using the ipv6 unicast-host-routecommand. You can also define IPv4 /32 route prefixes in
the host table using the ipv4 unicast-host-routecommand.
310
Equal Cost Multi-Path (ECMP)