ODUflex Overview:
New to hierarchy (Oct-09)
Two flavours of ODUflex standardization:
Circuit ODUflex
– Supports any possible client bit rate as a service in circuit transport networks
– CBR clients use a bit-sync mapping into ODUflex (239/238xthe client rate)
Packet ODUflex:
– Creates variable size packet trunks (containing GFP-F mapped packet data) for transporting packet flows using L1 switching of a LO ODU
– In principle, can be of any size, but in a practical implementation it will be chosen to be multiples of the lowest tributary slot size in the network
Similar to VCAT, but avoids differential delay problem by constraining the entire ODUflex to be carried over the same higher order ODUk, and provides one manageable transport entity per service (while also limiting the application to
ODUflex that fits within one higher order ODUk)
ODUflex Resizing:
Not yet defined in G.709; draft revision under discussion
Mixed operator views
Component vendor hesitation due to fear of delay
Some vendor opposition.
While earlier agreement that hitless resizing should not be precluded, significant time pressure for establishing equipment/component “hooks” to enable resizing capability
Unlike VCAT/LCAS, impacts all equipment along path
Allay component vendor concerns re-investment payoff (LCAS was a significant development investment for almost no return)
Build understanding of benefits from an operator perspective (recognizing differing operator philosophies)
A coalition of system vendors and device manufacturers has made significant progress in specifying a technical solution that will be brought into standards after it is complete.
Generic Mapping Procedure (GMP):
Historically, G.709 had three multiplexing routes (ODU1→ODU2,ODU1 →ODU3, ODU2 →ODU3)
All options could be described explicitly with small numbers of complete fixed stuff columns
Two multiplexing routes are “clean” with a single justification opportunity per multi-frame
One multiplexing route is “messy” with justification opportunity location and spacing varying based on the particular tributary slots (TS) chosen
Revised G.709 adds over 100 new multiplexing routes (ODUflex into 1-80 TS of OPU4, 1-32 TS of OPU3, 1-8 TS of OPU2, ODU0 →ODU2,3,4, ODU2e →OPU3, ODU1,2,2e,3 →OPU4)
If a traditional justification approach is employed, almost all of the new multiplexing routes would be “messy” with number and location of justification opportunities varying according to the particular TSs assigned
Location of even “fixed” stuff would have to be determined algorithmically, because there are too many combinations to draw them all explicitly.
Offers possibility to support hitless resizing of packet ODUflex (precluded by traditional justification approach).
Generic Mapping Procedure (GMP)
Basic Concepts
Single mechanism used to accommodate the nominal bit-rate difference between the client and server, and the clock variations that may occur between client and server i.e., no distinction between “fixed” and “variable” stuff locations
The server frame (or multi-frame) is divided into a certain number of GMP “words”, where each word may contain either data or stuff.
Words containing data are distributed as evenly as possible (quantized to word size) across server frame using sigma/delta distribution algorithm
Correct operation depends only on mapper and demapper knowing the number of data words which are filled into each frame (or multi-frame)
Larger GMP word sizes are used for higher bit-rate clients to avoid the need for large barrel shifters in the implementation.
If necessary to meet the timing requirements of the client, additional timing information may be transmitted from the mapper to the demapper
Enables the demapper to know how many client bytes (or bits) are to be emitted by the demapper during each server frame period.
Applications
Client mappings into LO OPUk:
CBR clients less than OPU0 bit-rate into OPU0
CBR clients greater than OPU0 but less than OPU1 bit-rate into OPU1
CBR clients close to bit-rate of OPU2, OPU3, or OPU4 into the respective container
Note that CBR clients that are greater than OPU1 bit-rate but not a convenient fit for
OPU2, OPU3, or OPU4 are mapped via ODUflex
Tributary mappings into HO OPUk:
LO ODU0 into HO OPU2, OPU3
LO ODUflex into HO OPU2, OPU3
LO ODU2e into HO OPU3
Any LO ODUk (k=0, 1, 2, 2e, 3, flex) into HO OPU4
Comments are closed