It’s not too surprising to hear QoE experts and video delivery optimization providers, including DLVR, pushing multi-CDN solutions. But now even CDNs themselves are acknowledging the need to remove single points of failure with a multi-CDN strategy.
Let that sink in for a second. A CDN provider, who would necessarily lose some traffic and revenue, is suggesting that publishers should employ a multi-CDN strategy. As longtime providers in the streaming media ecosystem, we all understand that as streaming viewership grows, consumer expectations for higher QoE is growing as well, two trend lines at odds with each other. Buffering and service interruptions, not to mention total catastrophes like a live event outage, are setbacks for the entire industry. Improving our ability to collect finer points of streaming data to proactively and intelligently route video traffic is the path to eliminating these setbacks.
CDNs have matured rapidly since the early 2000s. There are now dozens of international and regional CDNs, plus a variety of hybrid in-house and cloud delivery options. Almost all of them perform well on average and in many situations. But there are simply too many variables impacting today’s video content models to rely on just one CDN. Each CDN has its own strengths and weakness, and while it’s not difficult to find a provider of CDN services that will do a fine job on average, most of the time it’s simply impossible to find a single CDN provider that will do the best job all of the time.
A multi-CDN approach allows content publishers more control over a variety of important factors including cost, regional performance, and most importantly, heading off a disaster when something goes wrong. The publisher that doesn’t have a multi-CDN strategy today is at risk of damaging their brand and losing viewers at any moment. While widespread outages are what make the news (sorry Amazon), the fact is that shifting conditions constantly impact any CDN’s ability to achieve its peak performance across its entire infrastructure. At a 30,000 foot view a CDN may seem to be doing a good job, but at any given moment somewhere a router port is congested, a server is overloaded, or a cache has missed on a less popular video request.
As with all technology there’s a range of approaches to a multi-CDN strategy. Randomly assigning requests distributes the load but doesn’t capitalize on each CDN’s changing performance. CDN selections based on test objects may leverage each CDN’s average performance at completing the test, but can’t measure and don’t utilize actual video streaming performance specific to each publisher. Full-featured, automated and intelligent solutions such as DLVR’s service, however, use real-time data to optimize each stream for peak performance, replacing randomized and/or averages-based processes with actual data and business rules-driven automation.
In side by side testing, one of our customers found that using DLVR’s delivery optimization provided a 77% decrease in traffic load times vs a CDN-only solution. In addition, when one CDN initiated throttling during peak viewing hours (the 9am hour and 7pm hour in local time zones), DLVR instantly detected the traffic slowdown and proactively switched the affected traffic to another CDN, preventing abandoned streams and delivering a higher average bit rate and a higher quality viewing experience.
For this publisher, without DLVR an estimated 670,000 users would have abandoned their streams due to slow load times in a single day. This really illustrates the quantifiable impact of proactive mitigation to counter changing CDN conditions.
Whichever path you take, some level of multi-CDN capability should now be considered a must-have feature of your streaming workflow. Consistently delivering higher levels of viewer satisfaction will drive the continued growth of streaming video and the digital transformation of the video industry.