Read the original LinkedIn article here: How Video Encoder Computing Efficiency Can Impact Streaming Service Quality
Written by:
Mark Donnigan is VP Marketing for Beamr, a high-performance video encoding innovation company.
Computer software is the bedrock of every function and department in the enterprise; accordingly, software video encoding is necessary to video streaming service operations. It's possible to enhance a video codec implementation and video encoder for two but rarely 3 of the pillars. It does state that to provide the quality of video experience customers expect, video suppliers will need to evaluate industrial solutions that have actually been performance enhanced for high core counts and multi-threaded processors such as those available from AMD and Intel.
With so much turmoil in the circulation design and go-to-market business plans for streaming entertainment video services, it may be tempting to push down the concern stack choice of brand-new, more effective software application video encoders. With software application consuming the video encoding function, compute efficiency is now the oxygen needed to thrive and win versus a significantly competitive and congested direct-to-consumer (D2C) marketplace.
How Video Encoder Computing Efficiency Can Impact Streaming Service Quality
Until public clouds and common computing turned software-based video operations mainstream, the procedure of video encoding was performed with purpose-built hardware.
And then, software ate the hardware ...
Marc Andreessen, the co-founder of Netscape and a16z the famous equity capital firm with financial investments in Foursquare, Skype, Twitter, box, Lyft, Airbnb, and other similarly disruptive companies, penned an article for the Wall Street Journal in 2011 titled "Why Software Is Eating The World." A variation of this post can be found on the a16z.com website here.
"6 years into the computer revolution, four years since the innovation of the microprocessor, and 2 years into the rise of the modern Web, all of the innovation needed to change industries through software lastly works and can be extensively provided at worldwide scale." Marc Andreessen
In following with Marc Andreessen's prophecy, today, software-based video encoders have nearly entirely subsumed video encoding hardware. With software application applications devoid of purpose-built hardware and able to work on ubiquitous computing platforms like Intel and AMD based x86 makers, in the data-center and virtual environments, it is entirely accurate to say that "software application is eating (or more appropriately, has actually consumed) the world."
What does this mean for a technology or video operations executive?
Computer system software application is the bedrock of every function and department in the business; appropriately, software video encoding is necessary to video streaming service operations. Software application video encoders can scale without requiring a direct boost in physical space and energies, unlike hardware. And software can be moved around the network and even whole data-centers in near real-time to satisfy capacity overruns or short-term rises. Software is much more flexible than hardware.
When handling software-based video encoding, the three pillars that every video encoding engineer should deal with are bitrate performance, quality preservation, and calculating efficiency.
It's possible to optimize a video codec execution and video encoder for two but hardly ever 3 of the pillars. Many video encoding operations hence focus on quality and bitrate performance, leaving the calculate performance vector open as a sort of wild card. As you will see, this is no longer a competitive technique.
The next frontier is software application computing performance.
Bitrate effectiveness with high video quality needs resource-intensive tools, which will cause slow operational speed or a considerable boost in CPU overhead. For a live encoding application where the encoder need to run at high speed to reach 60 frames-per-second (FPS), a compromise in bitrate efficiency or absolute quality is frequently needed.
Codec intricacy, such as that needed by HEVC, AV1, and the upcoming VVC, is outpacing bitrate efficiency developments and this has actually developed the need for video encoder efficiency optimization. Put another method, speed matters. Generally, this is not a location that video encoding professionals and image researchers have needed to be concerned with, but that is no longer the case.
Figure 1 illustrates the advantages of a software encoding implementation, which, when all attributes are normalized, such as FPS and objective quality metrics, can do twice as much work on the precise same AWS EC2 C5.18 xlarge instance.
In this example, the open-source encoders x264 and x265 are compared to Beamr's AVC and HEVC encoders, Beamr 4, and Beamr 5.
No alt text provided for this image
For services Mark Donnigan needing to encode live 4Kp60, one can see that it is possible with Beamr 5 however not with x265. Beamr 5 set to the x264 comparable 'ultrafast' mode can encode 4 private streams on a single AWS EC2 C5.18 xlarge circumstances while x265 operating in 'ultrafast' can not reach 60 FPS at 4K. As you can see in this poignant example, codec performance is directly related to the quality of service as an outcome of less devices and less complicated encoding frameworks needed.
For those services who are primarily worried with VOD and H. 264, the best half of the Figure 1 graphic shows the efficiency benefit of an efficiency optimized codec implementation that is established to produce really high quality with a high bitrate performance. Here one can see approximately a 2x advantage with Beamr 4 compared to x264.
Video encoding compute resources cost genuine money.
OPEX is thought about thoroughly by every video supplier. Suppose entertainment experiences like live 4K streaming can not be provided dependably as a result of an inequality between the video operations ability and the expectation of the customer.
Since of performance restrictions with how the open-source encoder x265 utilizes compute cores, it is not possible to encode a live 4Kp60 video stream on a single maker. This doesn't suggest that live 4K encoding in software application isn't possible. It does state that to deliver the quality of video experience consumers anticipate, video distributors will require to examine industrial services that have actually been performance optimized for high core counts and multi-threaded processors such as those readily available from AMD and Intel.
The requirement for software application to be optimized for higher core counts was recently highlighted by AMD CTO Mark Papermaster in an interview with Tom's Hardware.
Video suppliers desiring to utilize software for the flexibility and virtualization choices they supply will come across overly complicated engineering difficulties unless they select encoding engines where multi-processor scaling is belonging to the architecture of the software application encoder.
Here is a post that shows the speed benefit of Beamr 5 over x265.
Things to consider concerning computing performance and efficiency:
Do not go after the next more sophisticated codec without considering first the complexity/efficiency quotient. Dave Ronca, who led the encoding team at Netflix for 10 years and recently delegated sign up with Facebook in a comparable capability, just recently released an excellent article on the subject of codec complexity entitled, "Encoder Intricacy Strikes the Wall." It's tempting to think this is only a concern for video banners with 10s or hundreds of millions of subscribers, the same trade-off factors to consider need to be considered regardless of the size of your operations. A 30% bitrate savings for a 1 Mbps 480p H. 264 profile will return a 300 Kbps bandwidth savings. While a 30% cost savings at 1080p (H. 264), which is encoded at 3.5 Mbps, will provide more than triple the return, at a 1 Mbps savings. The point is, we need to thoroughly and methodically consider where we are investing our compute resources to get the maximum ROI possible.
A business software option will be developed by a devoted codec engineering team that can stabilize the requirements of bitrate performance, quality, and compute efficiency. Exactly why the architecture of x264 and x265 can not scale.
Firmly insist internal groups and specialists perform compute performance benchmarking on all software application encoding solutions under consideration. The 3 vectors to determine are absolute speed (FPS), individual stream density when FPS is held continuous, and the overall number of channels that can be produced on a single server utilizing a nominal ABR stack such as 4K, 1080p, 720p, 480p, and 360p. All encoders should produce equivalent video quality throughout all tests.
The next time your technical team plans a video encoder shoot out, ensure to ask what their test strategy is for benchmarking the calculate effectiveness (performance) of each service. With a lot turmoil in the circulation model and go-to-market company prepare for streaming entertainment video services, it may be appealing to press down the top priority stack selection of new, more efficient software video encoders. Surrendering this work could have an authentic effect on a service's competitiveness and capability to scale to satisfy future entertainment service requirements. With software application consuming the video encoding function, calculate performance is now the oxygen needed to prosper and win versus a progressively competitive and crowded direct-to-consumer (D2C) market.
You can experiment with Beamr's software video encoders today and get up to 100 hours of free HEVC and H. 264 video transcoding every month. CLICK ON THIS LINK
