Cisco Systems 1000 Series Manual De Usuario

over and begin affecting packet buffering on other MLP bundles. (The MLP lost-fragment timeout 
can be configured on the multilink virtual template interface using the ppp timeout multilink 
lost-fragment (seconds) (milliseconds) configuration command). 

By default, in MLPoLNS, the Cisco IOS XE software informs the MLP that packets may arrive out 
of order. This works well for upstream traffic, but does not address the order issue at the peer CPE 
device. The peer CPE device should also be configured to allow for receipt of out-of-order packets. 
In Cisco devices, this can be managed by configuring the ppp link reorders command at the bundle 
interface.

When the Cisco ASR 1000 Series Aggregation Services Routers function as both a PTA device and 
an LNS device simultaneously, locally terminated member links (PTA) and member links that are 
forwarded from the LAC are not supported within the same bundle.

The following restrictions apply to all variations of broadband MLP at PTA and LNS modes:

When defining an MLP bundle with multiple member-link sessions, we recommend that all the 
member-link sessions utilize the same physical interface or subinterface. If other broadband sessions 
are sharing the same interface, ensure that all the member-link sessions utilize the same physical 
interface or subinterface.

The following issues might occur because of splitting links across separate physical interfaces or 
subinterfaces:

MLP is a sequenced protocol and all the packets and fragments must be reordered and 
reassembled at the receiver, based on the MLP sequence number before the receiver forwards 
them. In such a scenario, packets traversing separate physical interfaces may cause additional 
packet latency disparity between links due to transmission delays and other issues associated 
with using multiple physical paths. The reordering and reassembly processing may require 
additional MLP buffering at the receiver.

MLP on the Cisco ASR 1000 Series Aggregation Services Routers performs congestion 
management of the MLP bundle based on the congestion state of the member-link sessions that 
make up the bundle. If member-links are distributed across multiple interfaces and sufficient 
congestion is detected in one or more member links, the bundle may be back pressured due to 
the congestion even if all the links in the bundle are not congested. By keeping all the links on 
the same physical interface or subinterface, the chance of back pressure due to one link being 
congested is reduced.

The Multilink PPP feature provides the load-balancing functionality over multiple WAN links, while 
providing multivendor interoperability, packet fragmentation, proper sequencing, and load calculation 
for both inbound and outbound traffic. Cisco implementation of MLP supports the fragmentation and 
packet-sequencing specifications described in RFC 1990.