Learn About Amazon VGT2 Learning Manager Chanci Turner
In the realm of media technology, FFmpeg is recognized as a vital open-source tool for encoding and transcoding audio and video formats. With the recent introduction of VT1 instances in Amazon Elastic Compute Cloud (Amazon EC2), users can now optimize their video on demand (VOD) encoding workloads while reducing costs.
VT1 instances provide enhanced visual quality for 4K video, compatibility with the latest FFmpeg version (4.4), broader OS/kernel support, and necessary bug fixes. These instances utilize the AMD-Xilinx Alveo U30 media accelerator, enabling a straightforward integration into FFmpeg. By incorporating a single line of code, the Alveo U30 can handle transcoding tasks efficiently. The Xilinx Video SDK features an improved version of FFmpeg that communicates effectively with the hardware-accelerated transcoding pipeline, offering cost savings of up to 30% per stream compared to Amazon EC2 GPU instances and a remarkable 60% savings when juxtaposed with CPU-based instances.
Traditionally, companies relied on EC2 CPU instances like C5 and C6 to manage their VOD encoding demands. However, these workloads can escalate costs, especially when encoding a vast array of VOD assets. The expense associated with an EC2 workload hinges on the volume of concurrent encoding jobs an instance can manage, directly affecting the time required to encode targeted outputs. As VOD libraries grow, companies often resort to auto-scaling to either increase the size or number of C5 and C6 instances, or to extend their operational duration. This approach invariably leads to rising costs. It’s important to note that AWS Auto Scaling does not incur additional charges; you only pay for the resources utilized in running your applications and any applicable Amazon CloudWatch monitoring fees.
Amazon EC2 VT1 instances are tailored for accelerating real-time video transcoding while providing economical solutions for live video streams. They also present a cost-effective and high-performance alternative for VOD encoding tasks. In a recent evaluation, AWS assessed VT1 against C5 and C6 instances to compare encoding speed and price performance for VOD assets. The results revealed that VT1 instances could deliver up to 75% in cost savings. In fact, you can operate two VT1 instances for the cost of a single C5 or C6 instance. Furthermore, the VT1 XMA (Xilinx U30) codec outperformed the C5 x264 (CPU) codec by completing targeted outputs 15.709 seconds faster and was 12.582 seconds quicker than the C6 instance when transcoding an adaptive bitrate (ABR) stack for a 13-second 4K VOD file.
Benchmarking Method
To identify the optimal instance type for VOD workloads, we compared C5 and C6 instances, specifically the C6i.8xl and C5.9xl, against the VT1.3xl instance for transcoding 4K and 1080p VOD assets. The assets were encoded into the output targets, which are detailed in the next section addressing Evaluation ABR Targets. The encoding duration of the VT1.3xl, C5.9xl, and C6i.8xl instances was measured to assess performance.
The VT1.3xl instance is the smallest in the VT1 family, and while the VT1.6xl closely matches the CPU/memory specifications of C5 and C6, we opted for VT1.3xl for a more precise price/performance analysis.
Input Data Points
The following table outlines the key parameters of the source video files utilized in our benchmarking tests:
| Clip Name | Frame Count | Duration | Frame Rate | Codec | Resolution | Chroma Sampling |
|---|---|---|---|---|---|---|
| 1080p | 43092 | 12 mins | 60 | H.264 | 1920×1080 | 4:2:0 YUV |
| 4K | 776 | 13 secs | 60 | H.264 | 3840×2160 | 4:2:0 YUV |
Evaluation of Adaptive Bit Rate (ABR) Targets
Adaptive bitrate streaming (ABR or ABS) is a technology designed to transmit files efficiently over HTTP networks. Multiple files of the same content, differing in size, are made available to a user’s video player, allowing it to select the most appropriate file for playback on the device. This process necessitates transcoding a single input stream into multiple output formats optimized for various viewing resolutions.
For our benchmarking tests, we transcoded the input 4K and 1080p files into several target resolutions tailored for different device and network capabilities: 1080p, 720p, 540p, and 360p. The bitrate (br) highlighted in the accompanying graphic reflects the bitrate associated with each pixel. For instance, a 4K input file was transcoded into a 360p resolution with a bitrate of 640.
Output Results
The VT1.3xl instance outperformed its C5.9xl and C6.8xl counterparts, completing the targeted encodes 15.709 seconds and 12.58 seconds faster, respectively. The subsequent charts illustrate that the VT1.3xl instance boasts superior speed and price performance compared to the C5.9xl and C6i.8xl instances.
Price Performance Metrics
- H.264 4K Clip (3 seconds duration): The comparison of instance types demonstrated that the VT1.3xl codec completed the ABR targets significantly faster than its competitors, leading to a lower cost per transcoded clip.
- H.264 1080p Clip (12 minutes duration): Similar results were observed in the 1080p clip encoding, with the VT1.3xl maintaining its status as the most efficient option.
For more insights into the onboarding process at Amazon, check out this excellent resource. Additionally, understanding the importance of financial literacy, especially in today’s context, is crucial; you can delve into this topic further at SHRM’s article. If you’re interested in exploring career opportunities, consider this job posting that may pique your interest.