The Streaming Architecture Guide
Stop guessing at bitrate ladders and CDN configurations. Master the protocols, codecs, and economics behind modern video streaming.
The Cost of Bad Encoding
Every unnecessary megabit costs money at scale. Suboptimal encoding profiles waste bandwidth, reduce device compatibility, and destroy the viewer experience.
Core Architecture
Protocols
Understand the shift from RTMP to HLS and DASH. Learn chunk sizing, low-latency CMAF, and manifest manipulation.
Read Protocol Guides →Codecs
AVC, HEVC, VP9, AV1, and VVC. Detailed analysis of encoding complexity, licensing realities, and device decode support.
Compare Codecs →Economics
The math behind build vs buy. Transcoding ROI, CDN commit tiers, and the true cost of Server-Side Ad Insertion (SSAI).
View Economic Models →Industry Standard Bitrate Ladders (1080p Target)
A modern per-title encoding baseline for complex content (Sports/Action) vs simple content (Talking heads).
| Resolution | FPS | AVC (H.264) Bitrate | HEVC/VP9 Bitrate | Use Case |
|---|---|---|---|---|
| 1080p | 60 | 6.0 - 8.0 Mbps | 4.0 - 5.5 Mbps | Premium Tier / Action |
| 1080p | 30 | 4.5 - 6.0 Mbps | 3.0 - 4.5 Mbps | Standard Premium |
| 720p | 60 | 3.5 - 5.0 Mbps | 2.5 - 3.5 Mbps | Mid-tier / Mobile High |
| 720p | 30 | 2.5 - 3.5 Mbps | 1.5 - 2.5 Mbps | Mid-tier |
| 480p | 30 | 1.0 - 1.5 Mbps | 0.8 - 1.2 Mbps | Mobile Data Saver |
Encoding Efficiency Over Time
The historical progression of BD-Rate reduction from H.264 to VVC.
The Protocol War: HLS vs DASH
While RTMP dominated the flash era, modern streaming is built on HTTP. Apple's HLS (HTTP Live Streaming) and MPEG-DASH dominate the landscape.
Historically, delivering both required doubling your origin storage because HLS required TS containers and DASH required fMP4. Today, CMAF (Common Media Application Format) allows a single set of media chunks to serve both manifests.
Read the Protocol GuideThe Codec Landscape
AV1
30% better than HEVC/VP9. Backed by AOMedia (Netflix, Google, Apple). The future of web video as hardware decoding rolls out.
Read more →VVC (H.266)
Next-gen standard aimed at 8K and immersive video. Currently too computationally expensive for mainstream web deployment.
Build vs Buy: The Economics
SaaS encoding APIs (AWS MediaConvert, Mux) charge per minute. If you run a 24/7 linear FAST channel, SaaS pricing will destroy your margins within weeks.
Understanding when to rack your own hardware (ASIC/GPU transcoders) versus when to leverage cloud scalability is the most important financial decision in streaming architecture.
Engineering Calculators
10+ Interactive, client-side tools to model your streaming architecture. No server round-trips required.
Infrastructure & Cost Tools
Video Optimization Tools
Mastering CDN Caching
Live video wreaks havoc on CDNs if TTLs (Time to Live) are set incorrectly. Segments (.ts, .m4s) are immutable and should be cached for a year. Manifests (.m3u8, .mpd) change every few seconds and require micro-caching.
Failing to cache properly leads to thundering herd problems that will crash your origin server during a live event.
Multi-DRM Strategy
Protecting premium content requires navigating the fragmented DRM ecosystem. You need Widevine (Google/Android), FairPlay (Apple/iOS), and PlayReady (Microsoft/Windows).
Using Common Encryption (CENC), you can encrypt your media once and attach multiple DRM licenses to the same manifest.
Architecture FAQ
Direct answers to frequent debates among video engineers.
SRT is unequivocally better for modern ingest. It handles packet loss via ARQ rather than TCP's windowing, meaning it survives unstable first-mile connections significantly better than RTMP. It also supports HEVC out of the box, whereas RTMP requires proprietary extensions.
For high-tier VoD, yes. YouTube and Netflix deploy it aggressively. For live streaming, hardware encoding is required to keep latency and costs viable. Decode support is now standard in modern Android devices and newer Apple silicon.
Standard HLS dictates a client buffers 3 chunks before playback. If your chunk duration is 5 seconds, baseline latency is 15 seconds. To fix this, you must adopt Low-Latency HLS (LL-HLS) using partial segments and chunked transfer encoding.
Ready to optimize your streaming stack?
Dive into the technical guides or start modeling your infrastructure costs.
View All Guides