The rise of Mile High Video (MHV), from a half-day DASH workshop in 2016 to a three-day pilgrimage for video engineers, is best measured by the technologies it has spawned. Common Media Client Data (CMCD) is arguably its greatest offspring, a standard born from an industry daydream at MHV 2018, where video players and CDNs would finally speak the same language.
Fast forward to 2020, and CMCD became official, courtesy of the Consumer Technology Association (CTA). By 2024, the protocol had hit puberty—Apple’s AVPlayer joined the ranks, DVB signaled intent to integrate, and whispers of CMCD 2.0 surfaced, courtesy of Akamai. Now, at MHV 2025, Akamai’s Chief Cloud Architect Will Law is filling in the gaps, detailing the evolution from a promising protocol to a system fit for prime time.
The first iteration of CMCD assumed most players wanted three segments in the buffer before starting playback—a crude approximation at best. CMCD 2.0 sharpens that metric, introducing Media Start Delay, explicitly measuring the lag between a user pressing play and actual video appearing. A Start-Up Indicator flags when buffering transitions to playback, further refining diagnostics.
The telemetry payload has also been overhauled. Traditional CMCD latched metadata to requests, but CMCD 2.0 introduces Response Mode, a post-download reporting mechanism that captures first-byte and last-byte timings. More importantly, this mode liberates analytics from the CDN, allowing third-party monitoring services to collect it independently.
CMCD 2.0 is also gunning for advertising visibility—something that is a historically tricky beast. By tracking player state transitions (pre-roll, primary content, mid-roll), the standard aims to provide precise playback insights, including encoded bitrate drops and even proposals for VMAF-based quality assessments.
The specification also introduces standardized error codes, borrowing from the SVTA’s common player error framework, offering a unified way to diagnose playback failures across different implementations.
Prefetching has also received an efficiency upgrade. The previous syntax for requesting multiple assets was clunky, so CMCD 2.0 streamlines it, removing unnecessary elements like Next Request to improve performance.
While CMCD tightens playback analytics, a separate challenge looms in CDN failure management. Synamedia’s Gwendal Simon, CDN Architect, argues that recovering from an outage is easy, but rebuilding trust in a CDN is the real struggle. A proposed solution is Content Steering, a system that monitors CDNs at a regional level and dynamically shifts traffic to healthier alternatives.
Two approaches exist:
- Probe-Intensive Recovery – Switch users away, run diagnostic probes, and restore traffic upon positive reports. This is accurate, but computationally expensive.
- User-Sourced Feedback – Redirect users, have them stream from the failing CDN, and report performance metrics. This is less intrusive but sacrifices some user experience.
To mitigate impact, Synamedia’s Simon proposes CMCD’s buffer level (bl) key, using real-time buffer depth to determine how much quality degradation is tolerable before switching. The result is apparently a more scalable approach that doesn’t require direct player integration but does introduce trade-offs between QoE and service resilience.
Also on the CMCD train is Fraunhofer Fokus’s take on the latest iteration of dash.js, version 5, presenting at MHV how CMCD can integrate with modern players. We hear that dash.js v5 is primed for CMCD session initialization, aligning with the 6th edition of the MPEG-DASH standard.
Future enhancements include server guided ad insertion (SGAI) and low-latency SSRs for rapid startup times and seamless quality switches.
Yet, for all its advancements, CMCD still faces adoption hurdles. The need for multi-CDN analytics, cross-platform error tracking, and ad-tracking granularity remains high.
With CMCD 2.0 refining playback insights and expanding reporting capabilities, the final push must come from industry-wide standardization and real-world deployment.