Overview

Multilink PPP (MLPPP; also referred to as PPP Multilink, MLP, and MP) aggregates multiple physical links into a single logical bundle. More specifically, MLPPP bundles multiple link-layer channels into a single network-layer channel. Peers negotiate MLPPP during the initial phase of Link Control Protocol (LCP) option negotiation. Each router indicates that it is multilink capable by sending the multilink option as part of its initial LCP configuration request.

An MLPPP bundle can consist of multiple physical links of the same type—such as multiple asynchronous lines—or can consist of physical links of different types—such as leased synchronous lines and dial-up asynchronous lines.

The router acts on MLPPP like another PPP Network Control Protocol (NCP). Packets received with an MLPPP header are subject to fragmentation, reassembly, and sequencing. Packets received without the MLPPP header cannot be sequenced and can be delivered only on a first-come, first-served basis.

Application

Some users need more bandwidth than a T1 or an E1 channel can provide, but cannot afford the expense or do not need the bandwidth of T3 or E3. Equal-cost multipath (ECMP) is one way to achieve the desired bandwidth. MLPPP is commonly used as an alternative to ECMP to deliver NxT1 service. NxT1 service provides bandwidth greater than DS1 service without going up to the expense and infrastructure required for DS3 service. Cost-analysis of NxT1 versus DS3 service typically imposes a practical limit of 8xT1 service; that is, aggregation of no more than eight T1 or E1 connections into an MLPPP bundle.

The NxT1 implementation of MLPPP logically aggregates up to eight T1 or E1 connections into a single virtual connection, or bundle, to a given customer site, as shown in Figure 35.

Figure 35: MLPPP Aggregation of T1 Lines into a Single Bundle

Because MLPPP aggregates multiple link-layer channels onto a single network-layer IP interface, protocol layering within the router is different than for non-multilink PPP.

Figure 36 illustrates interface stacking with MLPPP.

Figure 36: Structure of MLPPP

MLPPP LCP Extensions

Multilink PPP adds the following LCP negotiation options:

• Multilink maximum received reconstructed unit (MRRU) option—The MRRU option has two functions. First, it informs the other end of the link the maximum size of the PPP packet payload that the router can receive. Second, it informs the other end that the router supports MLPPP. When you enable multilink on your router, the router includes the MRRU option in LCP negotiation with the value set to the maximum received unit (MRU) value for PPP. If the remote system rejects this option, the local system determines that the remote system does not support multilink PPP and it terminates the link without negotiation.

  Note: The router does not bring up a link if the MRU value received from a peer device differs from the MRRU value received from the peer.

• Short sequence number (SSN) header format option (not currently supported)—The SSN option indicates that the transmitting router wants to use a short sequence number (12 bits) in the MLPPP header rather than a long sequence number (24 bits). The router currently supports only long sequence numbers.
• Endpoint discriminator option—The endpoint discriminator option identifies the router transmitting the packet. If the receiving router determines that packets on another link have the same endpoint discriminator option, this link must be joined to that bundle. If the receiving router determines that no packets on other links have the same option, the receiving router must create a new bundle from this link.

  The endpoint discriminator is generated internally; you cannot configure it. The endpoint discriminator option is the same for all links on one end of the bundle; at the other end, all links also share a common endpoint discriminator. The two endpoint discriminators are different if the MLPPP bundle is set up between two E Series routers.

MLPPP Link Selection

By default, E Series routers use a round-robin algorithm to select the link on which to transmit data on an MLPPP interface. The round-robin link selection method applies to both best-effort packets, such as data, and non-best-effort (high-priority) packets, such as voice and video. Best-effort packets are encapsulated with an MLPPP header that contains a sequence number, whereas non-best-effort packets are encapsulated with a PPP header that does not contain a sequence number.

The member links in an MLPPP bundle can experience different queuing delays due to the volume of traffic transmitted on the MLPPP interface. These delays can cause packets to arrive out of order at the remote router. The effect of such delays differs for best-effort packets and non-best effort packets, as follows:

• For best-effort packets that arrive out of order from the E Series router, the remote router can use the sequence number to reorder and forward the packets in the correct order, regardless of the order in which the packets were received.
• For non-best-effort packets that arrive out of order from the E Series router, the lack of a sequence number prevents the remote router from being able to determine the correct order in which to forward the packets. This can cause problems with applications that require high-priority voice and video traffic transmitted on MLPPP interfaces to be received in the same order transmitted by the peer applications.

To ensure that the E Series router maintains the proper packet order when transmitting non-best-effort traffic, you can use the ppp hash-link-selection command to enable use of a hash-based algorithm to select the link on which the router transmits high-priority packets on an MLPPP interface.

When you use hash-based link selection instead of the default round-robin link selection for non-best-effort traffic, the router uses the IP source address (SA) and IP destination address (DA) of the packet as a hash to select the MLPPP member link on which to transmit the packet. Specifically, the router uses the hash algorithm to bind the transmission of all traffic between this IP SA and IP DA to the same member link in the MLPPP bundle.

If the member link selected to transmit high-priority packets becomes inoperable or is removed from the MLPPP bundle, the router must select a different link on which to transmit the packets. As a result, packets transmitted on this new link might sometimes arrive at the remote destination before the traffic sent on the previously selected member link.

You can configure hash-based MLPPP link selection in any of the following ways:

• To configure hash-based link selection for a individual MLPPP member link, issue the ppp hash-link-selection command from Interface Configuration mode or Subinterface Configuration mode in the context of the individual link interface. For more information, see Configuring Static MLPPP.
• To configure hash-based link selection for all current member links in an MLPPP bundle, issue the ppp hash-link-selection command from Interface Configuration mode in the context of the MLPPP bundle. Doing this has the same effect as issuing the ppp hash-link-selection command separately for each member link in the bundle. For more information, see Contextual Command Differences.
• To configure hash-based link selection for all dynamic MLPPP link interfaces created by a profile, issue the ppp hash-link-selection command from Profile Configuration mode. For more information, see Configuring Dynamic MLPPP.

For a detailed description and examples of using the ppp hash-link-selection command, see ppp hash-link-selection.