Because there are only three EXP bits,
there are only eight possible combinations of output queue and drop precedence
that may be encoded into each packet.
?– L-LSP model In this model, QoS is inferred both from the label and the EXP bits
in the MPLS label shim. Each LSP has a dedicated queue assigned to it, and then
the EXP bits may be used to encode multiple levels of drop precedence within that
queue. This is an implementation of the hard QoS model because bandwidth may
be both reserved and policed per LSP. Since there are rarely more than three or
four levels of drop precedence, one EXP bit is ???spare??? and may be used for congestion
notification. In fact, it is also possible to accommodate congestion notification
in the E-LSP model by assigning codepoints to indicate the presence of congestion
??”but that, of course, further reduces the number of queues and/or drop precedences
that can be encoded in the EXP field.
The E-LSP and L-LSP models are defined in RFC 3270 (???MPLS Support of
Differentiated Services???), though there has been little or no implementation of the
signalling extensions defined in that RFC.
MPLS Applications
Having discussed the MPLS technology it is useful at this point to briefly introduce
some of the applications for MPLS.
Pages:
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979