By Juan Martin
Are we surprised? A legendary boxer and a top digital influencer attracting record high viewers across territories, across platforms, uniting cross generational audiences in an unprecedented celebration of sports and entertainment. Jake Paul and Mike Tyson were poised to headline one of the most talked-about live-streaming boxing matches ever. But was the world’s largest streaming platform, Netflix, ready for it?
Instead, the hashtags “#Unwatchable” and “#NetflixCrash” is what made the news last Friday night, as the streaming giant struggled to put on the show it had promised fans. Over 500,000 users reported technical issues during the event according to DownDetector.
Netflix is new to live programming. It’s an attempt to expand beyond movies, television shows and documentaries to engage and retain its now close to 300 million subscribers. Yet, as it became obvious on Friday night, Video on Demand (VOD) is not the same as live streaming. Live events, especially live sports, place very different demands and scale on the technology stack:
- You have to encode and distribute content in (close to) real time, which doesn’t allow for strategically loading caches around the world as you would do in VOD
- Users watch the fast-paced action on big screens that demand not only high definition video but high frame rates, which increases the payload
- Traffic comes in spikes, particularly when the event starts and flash crowds rush in. This is even more critical in pay-per-view events as there are payments to be made and entitlements to check
- And finally, you are distributing over the Internet, which is an unmanaged network with harder to predict traffic/capacity in a complex mix of peering networks, regions and ISPs
It’s hard to pinpoint exactly what went wrong with Netflix’s live streaming event from an outside perspective, but here are a few things we do know. They are working to adapt an architecture that was originally designed and optimized for VOD, not live streaming. As they evolve the platform, issues may only become apparent until the platform is under stress. This was the case in 2023 during their first major live event, the “Love Is Blind” Reunion, when a software bug went undetected, leading to the cancellation of the live stream.
We also know they rely on a single, proprietary CDN (Open Connect) and, when caches run out and traffic faces constraints in the peering network, there’s no fallback. If combined with higher-than-anticipated peak traffic, some ISPs may not be able to access the content and impact the end-user experience.
More importantly, we know that live sports events are hard and they require industry experience, careful planning, and fine-tuning over many cycles to adapt to user and service patterns. Here are a few examples of how we help our customers prepare:
- Model out peak infrastructure needs and build redundancy where possible, e.g., multi-CDN strategy, auto-scaling resources
- Design overlapping caching layers to push distribution towards the edge of the network and protect origin servers
- Stress-test the platform ahead of high-concurrency events, mirroring your audience’s behavior and regional dispersion
- Build fail-safes for authentication, authorization, and payment gateways to avoid bottlenecks during ramp up periods
- Have trained specialists on-site and on-call to provide real time monitoring and proactive incident management
Netflix is probably trying to learn from this experience, as they prepare for their NFL debut on Christmas Day (they have two games scheduled), and the upcoming WWE “Raw” in early January. Time will tell how Netflix fares with these live events or live events in general!
How are you getting ready for your next live sports event?
