I described what Adaptive Bit Rate Streaming (ABR) is in my prior article, if you need that background. Also there is an ABR Lab Exercise if you want some hands on learning with ABR that will show you visually some of the items discussed below.
Here, I wanted to dive a little deeper into this protocol because to fully understand streaming performance, you must look beyond ABR alone and examine how it interacts with the transport layer. As we all know, the transport layer is UDP, TCP and QUIC. TCP and QUIC are not simple protocols. ABR is an application-layer control loop that depends heavily on the behavior of the underlying congestion control algorithm.
Two Interacting Control Loops
There are two independent but tightly coupled systems:
| Layer | Control Mechanism | Input Signal | Output Action |
|---|---|---|---|
| Transport | Congestion Control (CUBIC, BBR, QUIC CC) | Loss, RTT, bandwidth | Adjust send rate |
| Application | ABR Algorithm | Throughput, buffer level | Adjust requested bitrate |
The key insight is that streaming performance is the result of how these two loops interact.
1. ABR over TCP CUBIC (Loss-Based Control)
TCP CUBIC is the dominant TCP congestion control algorithm in most Internet deployments. TCP CUBIC became the dominant congestion control algorithm largely because it strikes a practical balance between scalability, fairness, and deployability across modern high-speed Internet paths—especially in Linux-based infrastructure that underpins most servers and CDNs. To read more on TCP CUBIC – look here.
TCP CUBIC Behavior Characteristics
- Aggressively increases throughput
- Uses packet loss as the primary congestion signal
- Reduces rate sharply upon detecting loss
The impact that TCP CUBIC has on ABR, at least from the ABR perspective, is that throughput ramps up quickly, packet loss causes sudden drops, and retransmissions introduce delay. The resulting User Experience is a possible bitrate oscillation (up/down switching), buffer instability, and reduced overall video quality in lossy environments.
Now if you are a Wireshark packet analysis person, you will need to pay attention for TCP retransmissions and duplicate ACKs, you will see a sawtooth throughput pattern and variable segment download times.
2. ABR over TCP BBR (Model-Based Control)
TCP BBR takes a fundamentally different approach.
TCP BBR Behavior Characteristics
- Models bottleneck bandwidth and RTT
- Avoids filling queues unnecessarily
- Minimizes packet loss
The impact that TCP BBR has on ABR, at least from the ABR perspective is more stable throughput, with fewer retransmissions and more predictable delivery timing. The resulting User Experience is a stable bitrate selection, fewer quality drops and likely higher sustained resolution.
Again, if you are a Wireshark person, you will see minimal retransmissions,a smooth throughput curve, and subtle pacing cycles
3. ABR over QUIC (Modern Transport Evolution)
QUIC, used in HTTP/3, changes/affects the transport layer significantly.
QUIC Key Architectural Advantages
- Runs over UDP (bypasses TCP limitations)
- Integrated encryption (TLS by default)
- Eliminates head-of-line blocking
- Supports user-space congestion control (often BBR-based)
The impact that QUIC has on ABR, at least from the ABR perspective is faster startup times (reduced handshake latency), improved performance under packet loss, and a more consistent segment delivery. The resulting user experience is faster initial playback, improved quality stability on Wi-Fi and mobile networks, and better resilience to transient loss.
Wireshark folks – QUIC uses UDP-based traffic instead of TCP, and it has encrypted payloads (thus limited visibility), but consistent packet pacing.
The Critical Engineering Insight: Feedback Mismatch
The primary challenge in streaming performance is feedback misalignment between layers:
- CUBIC + ABR
- Loss interpreted as congestion
- ABR overreacts → bitrate drops
- Leads to oscillation
- BBR + ABR
- Stable delivery → accurate bandwidth estimation
- ABR decisions improve
- QUIC + ABR
- Faster feedback loop
- Better alignment with application requirements
Buffer Dynamics: The Real Decision Engine
ABR decisions are heavily influenced by buffer occupancy:
- In CUBIC environments:
- Loss events drain the buffer
- ABR aggressively reduces bitrate
- In BBR/QUIC environments:
- Buffer remains stable
- ABR can safely increase bitrate
This is why modern ABR algorithms (such as buffer-based approaches) perform significantly better with BBR and QUIC.
Some Practical Implications for Broadband and Wi-Fi Networks
Wi-Fi Environments
- Loss from interference is misinterpreted by CUBIC as congestion – there will be a tendency to reduce quality of the video as the congestion control backs off.
- BBR and QUIC handle these conditions more effectively
High-Latency Links
- CUBIC struggles with slow ramp-up
- BBR and QUIC improve startup performance
Speed Testing vs Real Applications
- Traditional TCP-based speed tests do not reflect streaming behavior – this is not good
- QUIC-based applications behave differently than TCP flows and as we have discussed in prior QUIC articles, because QUIC is encrypted, this is hard to diagnose in packet captures
Network performance is not just about throughput—transport behavior directly shapes application experience.
My Summary
Adaptive bitrate streaming cannot be fully understood without analyzing the transport layer beneath it.
- TCP CUBIC introduces variability due to its loss-based model
- TCP BBR improves stability through bandwidth modeling
- QUIC modernizes transport with faster, more resilient delivery
Together, they form a layered system where congestion control and application adaptation are tightly coupled, and mastering that interaction is essential for accurate troubleshooting and performance optimization.
Comments are welcomed below from registered users. You can also leave comments at our Discord server.
If you would like to see more content and articles like this, please support us by clicking the patron link where you will receive free bonus access to courses and more, or simply buying us a cup of coffee!