Feasibility of TE mechanism in core networks

[Florian-Daniel Otel <otel@ce.chalmers.se>]

Dear all,


I have the following problem regarding the feasiblity of a TE
mechanism and I'm looking for feedback from ppl w/ operational
experience/insight:

How reasonable/feasible to implement would be a TE mechanism that
would provide hard guarantees (i.e. worst-case bounds) for the queuing
delay and the amount of in-flight data for delay-sensitive traffic
but at the cost of following restrictions: 

    - There is a dedicated infrastructure of links for carrying this
type of traffic with all links of equal capacity, parrallel links
allowed (e.g. dedicated lambdas). Each link is driven by a dedicated
work-conserving FIFO queue. 
    - Trraffic trunks are carried w/ from ingress to egress  with one ER-LSP
per trunk i.e. no tunneling, label merging, etc. Each trunk is strictly
rate limited and ingress policed/shaped. 
    - All packets have a fixed size (!). Not "maximum" but
"fixed"  e.g. carrying G.729 voice hardware encoder that is generating
fixed sized payloads of 20 bytes. 

Putting it simply, an ATMish core network setup using dedicated links
of equal capacity. 

How does this picture sound from an (educated) operational perspective ?

Thanks in advance,

Florian