The day the Internet died
The inauguration of the 44th president was to be a highlight event for net weenies. They expected to enjoy full HD, full-screen streaming video right to their office or home PC. Tens of millions of people expected to watch the day’s events on their desktops. Why bother with old-fashioned broadcast channels? Let the Web deliver the news, right?
Wrong.
While many net users perceived the event as the culmination of net over broadcast, when it came time to deliver the goods, the Internet failed miserably. The cause for that failure is something with which we broadcasters are quite familiar. It’s called bandwidth.
It matters not whether there is one viewer or there are 1 billion viewers, a single TV transmitter can theoretically reach every one of them and never burp. It doesn’t cost one penny more or require one extra hertz of bandwidth to reach as many people as you want because it is broadcast. That old-fashioned one-to-many communication model many think died in 1999 when Al Gore created the Internet still works beautifully.
I just had to laugh as I reviewed my familiar blog sites on the Wednesday morning after the much-publicized event. The blogs were replete with complaint after complaint about “video frozen,” “stuttering audio,” “live streaming failed” and other nonprintable words describing viewers’ experience in trying to watch the inauguration.
While trying to use the Internet for search functions during the day, I discovered my typically fast connection of 10Mb/s had slowed to a crawl of 25kb/s. The whole Internet performance was pretty much worthless all of Tuesday. I finally just gave up and worked on another project.
Web viewers discovered that the Internet starts to fail when simultaneous streaming reaches about 1 million. While that may be a lot of viewers for the Internet, almost any broadcast show reaches many times that number of viewers without any problem. Broadcast still works.
As the bloggers whined and moaned that their feeds were broken and unwatchable, those viewers watching on their TV sets saw brilliant HD video with stereo audio all in living color. Even though the Internet providers crowed about the numbers of streams they provided, each failed to mention that those streams were often unstable, pixilated, missing audio, stuttering and typically gave PC viewers little more than the message, “Waiting to connect.”
January 28th, 2009 at 10:42 am
Brad,
Great article that hits on a couple of key points about video over the Internet.
The way that most video get to end users over the Internet is through the use of Content Delivery Networks (CDNs) like Akamai, Limelight, Level 3, CDNetworks, and many others. These CDNs have peering relationships with the major interchanges to help push content that is in demand to end users. The Presidential Inauguration pushed the limits of a few channels; the CDNs capability to handle live streams, peering relationships, and the last mile of the ISPs and MSOs to deliver the content. The CDNs themselves have enough bandwidth available to support major events such as Major League Baseball Games, Victoria Secret Events, and the like. These events are too expensive for the content owners to provide delivery on their own so they leverage the extensive capabilities of CDNs and rent their networks to deliver their content.
If we think of the number of subscribers in traditional terms in the broadcast environment we often turn to Nielsen ratings in terms of 1 million households equaling one Nielsen point. In the Internet Video world, there isn’t the same measurement capabilities, but if we take the same measurement for a show like CSI, with some 25 million viewers, or 25 Nielsen points and convert it to an Internet Video event, that is some 25 million viewers simultaneously watching a stream that will likely be at 750Kb/s to 1Mb/s, let’s call it 1Mb/s for ease of calculations. At 25 millions streams at 1Mb/s we are talking about 25 million Mb/s or 25Tb/s of traffic. I believe Akamai’s network is concurrently provising some 4+ Tb/s for all content that they deliver. Think about it for a minute, one show could easily exceed the largest CDN’s capability to provide delivery of a single show online. Thankfully, most of the audience tunes in and watches a Video On Demand stream at different times of the week, usually the day after the show has its initial broadcast, or right before the new episode hits the next week.
What I am getting at is how easy it is to see how the push to online video delivery will easily push the networks and providers beyond their current capabilities and newer technologies will be required to deliver the content and store the content - another part of the discussion that I have not mentioned is the backend storage and delivery technology to keep up with these demands - another generation of technologies will be needed there that automatically adapts to the real-world viewing of the audience and adjusts resources according on the back-end providing just in time delivery so content that is highly in demand gets its quality of service and content that is hitting the long tail gets it right amount of resources.
Thanks for letting me share my thoughts.
Brett
Leave a Comment