6.3.1 E-LSP

E-LSPs can exist over link layer technologies in which MPLS shim layer is used, since the PHB Scheduling Class (PSC) is inferred from the EXP field in the MPLS shim layer header (refer figure 2.1). Since the EXP field is 3 bits, the E-LSP supports 8 BAs for a given FEC, irrespective of how many OAs these BAs belong to. The EXP-to-PHB mapping is either explicitly signaled at label set-up or is done by manual configuration. Since the 6 bit DSCP define a maximum of 64 PHBs, while the 3 EXP bits in MPLS header define a maximum of 8 values, the ingress LSR may have to maintain a mapping of up to 64 PHBs to 8 EXP values. It is important to note that the EXP not only conveys the PSC but also the DP incase of AF classes.

6.3.1.1 Benefits and limitations

The benefits provided by the use of E-LSPs are [13]:

1. If the EXP-to-PHB mapping is done manually, then E-LSPs can be supported by the existing signaling and label-distribution protocols without modification.

2. Support of multiple PHBs over a single LSP simplifies network management and also preserves the label space

There are also certain limitations that exist with the E-LSP usage, which are as follows [13]:

1. Since the E-LSP needs an EXP field which is specific to MPLS shim layer header, E-LSPs cannot be supported by link layer technologies like ATM and FRR.

2. During preemption, the lower priority E-LSP will be preempted irrespective of whether the E-LSP supports EF PHB or not and hence even the premium services through that LSP will be affected.

3. The TE decisions for setting up of an E-LSP are more complicated since the path chosen has to satisfy the constraints of an aggregate FEC containing multiple PHBs rather than choosing an optimal path for satisfying the constraints of a single PHB.

4. And last and most obvious point is that the E-LSP cannot support more than 8 PHBs and hence to fulfill that gap a new approach of determining the PSC from the MPLS label itself, is used.

6.3.2 L-LSP

An L-LSP is established for a single <FEC,OA> pair, i.e. the label assigned to a L-LSP not only determines the FEC, and hence the forwarding decision, but also determines the particular PHB to be applied to the packet during the forwarding decision. The EXP field, if it exists, is used for determination of the DP. In case of ATM, DP is carried within the CLP bit.

With such LSPs the PSC is explicitly signaled at the time of label establishment, in order for the label to carry both the FEC and PHB information. Hence, to support the L-LSPs the signaling protocols have been extended. For example the RSVP-TE signaling protocol has been extended to carry a DiffServ object for L-LSP as well as E-LSP.

6.3.2.1 Benefits and limitations

From the above discussion, it is easy to see the benefits, which are as follows [13]:

1. In comparison to E-LSPs which allow only 8 PHBs per FEC, L-LSPs permit 64 PHBs.

2. The L-LSPs does not require MPLS shim layer header, and hence can be supported by any link layer.

3. L-LSP allows computation of TE paths optimal for a particular service class.

Although there doesn;t seem to be any major drawbacks/limitations with L-LSPs, but it is not difficult to envisage the network management problems, in case of set up of a large number of LSPs for different application requirements.

The network management problems can be handled effectively with the use of network management protocols like Simple Network Management Protocol (SNMP), which is the object of discussion of the next chapter.