Summary
Non-traditional trade-offs will be required for good/optimal design of real-time (voice-on) IP applications
- for example, for delay measurements to be useful for voice on IP applications it will be helpful to know the entire delay distribution so app can use trade-offs
How does one obtain and represent end-to-end delay distributions, and what level of granularity is needed?
What is the distribution of the packet loss process itself?
Impact of Self-similarity?