The documentation set for this product strives to use bias-free language. For the purposes of this documentation set, bias-free is defined as language that does not imply discrimination based on age, disability, gender, racial identity, ethnic identity, sexual orientation, socioeconomic status, and intersectionality. Exceptions may be present in the documentation due to language that is hardcoded in the user interfaces of the product software, language used based on RFP documentation, or language that is used by a referenced third-party product. Learn more about how Cisco is using Inclusive Language.
This feature provides the capability to load balance traffic proportionally across multiple paths, with different cost.
Prior to this release, the higher bandwidth links used to carry the same traffic as the lower bandwidth links were underutilized.
Use the following new command to configure local Unequal Cost Multi Path (UCMP):
ucmp localprefix-listprefix-list-name
Load balancing is a forwarding mechanism that distributes traffic over multiple links
based on certain parameters. Prior to Cisco IOS XE Cupertino Release 17.8.1, the traffic
with different bandwidths could not be load balanced. Configuring same metric would make
higher bandwidth links carry the same traffic as the lower bandwidth links. Hence, the
higher bandwidth links were underutilized.
Generally, higher bandwidth paths have lower Interior Gateway Protocol (IGP) metrics
configured, so that they form the shortest IGP paths. Starting with Cisco IOS XE
Cupertino Release 17.8.1, with the Unequal Cost Multi Path (UCMP) load balancing
enabled, protocols can use even lower bandwidth paths or higher cost paths for traffic,
and can install these paths to the Forwarding Information Base (FIB). This feature is
only supported on Intermediate System to Intermediate System (IS-IS) protocol.
Starting with Cisco IOS XE Dublin 17.12.1, UCMP is supported on ASR 920 Routers
.
Starting with Cisco IOS XE Cupertino 17.8.1, UCMP is supported on
.
Consider the above topology where there are nine 10G POS links from A1 to C1 and three
40G gigabit ethernet links from A1 to C2. If you want to load balance traffic from A1 to
A2, across all the links from A1 to C1 and A1 to C2, you must configure equal metric on
all the links. This will create Equal Cost Multipath (ECMP) paths. However, the higher
bandwidth links will carry the same traffic as the lower bandwidth links and the higher
bandwidth links are underutilized. More specifically, if 9X10G POS links together carry
a load of 90 units, then the 3X40G links should also carry a “load” of 90 units, though
they have a capability of carrying a total load of 120 units.
To avoid this problem, UCMP allows you configure all the links to distribute the traffic
proportionately across the links based on bandwidth, even if the configured metrics on
all links are the same. With the UCMP load-balancing enabled, the protocols still
install multiple paths to the same destination in FIB, but each path will have a 'load
metric/weight' associated with it. FIB uses this load metric/weight to decide the amount
of traffic that needs to be sent on a higher bandwidth path and the amount of traffic
that needs to be sent on a lower bandwidth path.
In the following example, there are three paths to reach Network X as follows:
Paths
Cost from Router E to Net -X
E-B-A
40
E-C-A
50
E-D-A
40
IGP selects the lowest path links, i.e E-B-A and E-D-A. The path E-C-A is not considered
for load balancing because of higher cost. The lowest path link E-D (5) is not a tie
breaker, as the end to end cost to the Network X is considered.
Restrictions
UCMP is supported only by the IS-IS protocol.
IPv6 UCMP is not supported by the IS-IS protocol.
Native UCMP is not supported.
UCMP is not supported in L2VPNs.
In L3VPN, BGP Prefix-Independent Convergence (PIC) core is enabled by default. When the number of prefixes is more than six
and UCMP load balancing is converted into Indirect load balancing, ECMP and UCMP do not work. We recommend that you disable
BGP PIC Core for UCMP load balancing.
The router supports only 6K prefixes when UCMP is enabled. This includes both IPv4 and VPNv4 prefixes.
The packet processor has a maximum limit of 8 buckets to load-balance for each load balance ID. It is applicable for the Cisco
IOS XE RSP2 module.
ECMP vs UCMP
Equal Cost Multi Path (ECMP) is a forwarding mechanism for routing packets along multiple
paths of equal cost with the goal to achieve almost equally distributed link load
sharing. This significantly impacts a router's next-hop (path) decision.
UCMP applies a weight to a path. The weight applied is static and is
derived by the DMZ bandwidth extended community either assigned to a peer or as
configured via the Route Policy Language (RPL) route manipulation functionality.
In local UCMP, the best paths have the same metric. Though the metrics are same, IGP
calculates the 'load metric or weight' that is based on the bandwidth of each of the
links. This information is passed on to FIB and FIB takes care of load balancing the
traffic accordingly across the links.
Advantages of UCMP
This is a simple process.
No major changes are required in IGP other than obtaining the bandwidth of the
links. You can calculate local UCMP weights based on bandwidths and pass the
information to RIB or FIB.
Configure UCMP Load Balancing
To enable UCMP load balancing for IS-IS protocol per address family:
router isisucmp localprefix-listprefix-list-name
To enable local UCMP to calculate local weights based on bandwidths for IPv4 MT-0
routes:
Use show ip route to verify local UCMP configuration:
R1#show ip route12.12.12.12
Routing entry for 12.12.12.12/32
Known via "isis", distance 115, metric 60, type level-2
Redistributing via isis 1
Last update from 140.0.0.2 on BDI140, 00:00:09 ago
Routing Descriptor Blocks:
* 141.0.0.2, from 12.12.12.12, 00:00:09 ago, via BDI141, prefer-non-rib-labels, merge-labels
Route metric is 60, traffic share count is 1
MPLS label: 17012
MPLS Flags: NSF
Repair Path: 140.0.0.2, via BDI140
140.0.0.2, from 12.12.12.12, 00:00:09 ago, via BDI140, prefer-non-rib-labels, merge-labels
Route metric is 60, traffic share count is 10----UCMP Enabled,7:1 load share Expected
MPLS label: 17012
MPLS Flags: NSF
Repair Path: 141.0.0.2, via BDI141