VP9 Pricing: When Does Adding VP9 Lead to Bandwidth Cost Savings?
I’m just back from the Streaming Forum in London, where, during one of my sessions on codecs, a central question came up: When does it make economic sense to add a VP9 transcode to save bandwidth costs on the delivery of these streams to platforms that can play them? This is the classic breakeven analysis, in which you identify the fixed cost and divide it by the variable cost savings to determine the quantity necessary to recoup the fixed cost. In this column, I’ll explore this analysis regarding VP9. (Note that I’ve created and shared a spreadsheet that incorporates this analysis.)
The fixed cost in this case is encoding, which will vary according to your encoding tool and situation. For example, if you have boundless excess capacity, the cost could be zero, while if you’re paying a cloud encoder for the encoding, you’ll pay by the minute, or by the GB, or by the hour of encoding. Assuming you’re in the latter camp, the cost for encoding to VP9 format is dropping and is likely less than you’d think. For example, the Amazon Elastic Transcoder and Zencoder cloud encoders both support VP9 output and don’t appear to charge a premium.
An example on the Amazon CloudFront site calculates that a 10-minute source with 3 SD and 2 HD outputs, plus a single audio-only output, would cost $1.24 to encode. Let’s multiply this by six to get to an hour, and a total encoding cost of about $7.50. That’s our fixed cost. Obviously, the higher the fixed cost, the more streams necessary to break even, so if you’re performing this analysis, now might be a good time to see if you’re paying the lowest possible price for encoding.
The variable cost saving has two elements— the percentage bandwidth savings attributable to VP9 and the cost per GB for delivery. For the purposes of this analysis, I’m going to assume an average 37.5 percent bandwidth reduction, from 3.2 Mbps to 2 Mbps in my analysis.
Bandwidth costs are all over the map, and the less you pay, the more streams you need to deliver to offset the fixed cost. Looking at Dan Rayburn’s most recent CDN pricing analysis from May 2016, those spending more than $1M per year were paying between $0.005 and $0.008/GB, while those spending between $500K and $1M were paying from $0.007 to $0.012/GB. On the other hand, if you’re paying retail on CloudFront, you could be paying anywhere from $0.02 to $0.085/GB in the U.S., with prices often double or triple this for other regions.
At an encoding cost of $7.50, these delivery costs make a huge difference. If you’re paying $0.06/GB, you break even at 231 streams. If you’re paying $0.006, breakeven jumps to 2,315 streams. All these figures are available on the spreadsheet referred to above, including computations for encoding costs ranging from $3 to $20/unit.
Some final points: First, this analysis is not all about cost; it can be about capacity within a fixed pipe or quality of service. For example, JW Player started offering VP9 streams in mid-2016. According to SVP of product strategy John Luther, “65.29 percent of VP9 was viewed at 1080p, while only 28.21 percent of HLS video hit the optimum 1080p.” So, in addition to cost savings, VP9 improved the quality of the experience of those viewing these streams. For the record, Luther also reported that at 1080p, H.264 streams would have consumed 40 percent more bandwidth.
Also note that just because Amazon supports VP9 output doesn’t mean that it supports VP9 output in DASH format, which is what you’ll need for ABR delivery. So whichever encoder you use, make sure it can output DASH streams. And if you use DASH, you may need to add the new $0.05/DASH royalty into the breakeven analysis, adding a whole new level of cost and complexity.
This article appears in the April/May 2017 issue of Streaming Media magazine as "VP9: A Breakeven Analysis."
YouTube now endorses phones that provide the best streaming experience on its platform, and VP9 codec support is an essential.
Sometimes an industry benchmark—such as Netflix's ISP Speed Index—only tells half the story about video delivery. Before picking a codec, dig a little deeper.
Netflix compared 5,000 clips from 500 titles in its library using the x264, x265, and libvpx codecs. x265's implementation of HEVC was the clear winner on quality and efficiency, but whether that matters in light of compatibility and licensing issues isn't so obvious.
Publishers and encoding companies alike are beginning to embrace VP9, Google's open source codec. Here's how it stacks up on quality and data rates.
VP9 is the open-source codec from Google, and provides a royalty-free alternative to HEVC. It's more efficient than H.264, and while it's less efficient than HEVC, it compares well on quality.
HEVC might be getting all the attention, but Google's VP9 appears to match H.265 in quality and might play a bigger role in the transition from Flash to HTML5