Episode

Episode 862 - Have Your CAKE and Eat It Too

Podcast: FLOSS Weekly
Published: Jan 28, 2026
Duration seconds: 4050
Processing state: processed
Canonical source: https://b27a1a7a-dbb9-4922-ac3e-601e2d6ca1e1.libsyn.com/episode-862-have-your-cake-and-eat-it-too
Audio: https://traffic.libsyn.com/secure/b27a1a7a-dbb9-4922-ac3e-601e2d6ca1e1/FLOSS-862.mp3?dest-id=4272468
JSON: /v1/public/podcasts/floss-weekly/episodes/episode-862-have-your-cake-and-eat-it-too
Markdown: /podcast/floss-weekly/episode-862-have-your-cake-and-eat-it-too.md

Actions

POST https://stenobird.com/v1/public/podcasts/floss-weekly/episodes/episode-862-have-your-cake-and-eat-it-too/transcription-requests
Idempotently request low-priority transcript generation for this episode.
GET https://stenobird.com/podcast/floss-weekly/episode-862-have-your-cake-and-eat-it-too.md
Read the agent-friendly Markdown representation of this episode resource.

Summary

Red Hat engineer Toke Hoiland-Jorgensen explains the mechanics of CAKE_MQ, a new Linux kernel innovation designed to eliminate bufferbloat. The discussion covers the evolution of queue management, the challenges of scaling networking across CPU cores, and the impact of modern traffic patterns on latency.

Topics

Linux Kernel
Networking
Bufferbloat
CAKE_MQ
eBPF
Red Hat
Queue Management
Latency

Highlights

Main idea: CAKE_MQ introduces a parallelizable approach to queue management to handle modern high-bandwidth, multi-core networking
Technical challenge: Maintaining global bandwidth enforcement across multiple CPU cores requires a single choke point, which can limit scalability
Failure mode: Traditional bufferbloat occurs when network bottlenecks create standing queues, leading to massive latency spikes during heavy usage
Practical takeaway: Modern video codecs and TCP flows are increasingly adaptive, scaling bandwidth usage in ways that interact with queue management
Observation: The Linux kernel's email-based workflow acts as a natural barrier against the influx of low-quality, AI-generated pull requests

Chapters

1:10 Introduction to Bufferbloat: An introduction to Toke Hoiland-Jorgensen and the fundamental problem of latency spikes caused by network bottlenecks.
5:55 The Mechanics of Good and Bad Queues: A deep dive into how queues absorb bursts and the distinction between beneficial buffering and harmful latency.
11:15 Adaptive Video Codecs and TCP: How modern streaming protocols scale bandwidth and how these fluctuations impact network congestion control.
16:20 CAKE and Connection Overhead: The technical implementation of CAKE, including handling DSL connections and calculating overhead compensation.
21:15 Scaling Networking Across Cores: The difficulty of enforcing global bandwidth limits when scaling network processing across multiple CPU cores.
26:15 The Trade-offs of Parallelism: Discussing the balance between high-performance scaling and maintaining fairness in pathological network cases.
41:20 BPF and the Future of Implementation: How eBPF provides a new way to implement networking algorithms and the challenges of low-level kernel debugging.