Computing delay (Di*) & backlog (Qi*) bounds
Given: connections 1,2,...,N; connection i
- has weight fi
- has input traffic that is (si, ri)-constrained (leaky bucket)
To determine the worst-case delay & backlog,
- determine a feasible order of the connections --- an order in which they reach zero backlog in the critical instance
- ?computational complexity?
- For each connection i in this order,
- detemine the latest instant at which its service rate is < ri
- ?computational complexity?