SAM

Evaluation

We evaluate SAM using topologies with a high heterogeneity of bandwidths and delays. The goal is to show that SAM has a small convergence time, remains stable in the optimal quality level, is fair with TCP, and allows receivers to use a rate proportional to the audience size of their session and the amount of bandwidth not used.

Inter-SAM Fairness and Bandwidth Utilization (Top 1):

Loss with increase of latency (topology only with one link with 1Mb/s and delay from 1 ms to 20 s):

SAM doesn't present anny loss when the latency increases.
Losses that should happen if receivers didn't know forehand the average rate of each layer .

We used delay above 1 s as a diagnostic tool to see the behavior in abnormal situations and to compare with the scenario used to evaluate RLM by Steve McCanne et al.

Convergence Time and Stability with four receivers (Top 2):

Thin granularity
Thick granularity
Convergence time if receivers didn't know forehand the average rate of each layer:

Convergence Time and Stability with incresing number of receivers (Top 3):

Fairness with TCP Flows (Top 4):