Bitrate Adaptation Algorithms (Rate vs. Buffer)

💡 The Core Idea

In DASH, the client must decide which bitrate to request for the next chunk ($R_{next}$) to maximize Quality of Experience (QoE).

🧠 Strategies

1. Rate-Based Adaptation

• Method: Estimates future bandwidth based on the throughput of the last few chunks.
• Flaw (Underestimation): In the “Steady State,” the client pauses (OFF period) when the buffer is full. This resets the TCP congestion window. When it restarts (ON period), TCP ramp-up is slow, causing the client to underestimate bandwidth and pick a lower quality than necessary.

2. Buffer-Based Adaptation

• Method: Selects bitrate based on buffer occupancy.
  • Low Buffer $\rightarrow$ Low Bitrate (to prevent stall).
  • High Buffer $\rightarrow$ High Bitrate.
  • $R_{next} = f(buffer_{now})$.
• Flaw: Can oscillate unnecessarily between bitrates if the steps are too aggressive.

🔗 Connections

• Source: Source - Multimedia Applications