table of Contents
Article Directory
- table of Contents
- Foreword
- Traditional self-built video systems issue
- Video Cloud (VaaS) the definition and origin
- Video Cloud application scenarios
- Video Cloud Market
- The principle of cloud video
- Video Cloud Case
- Video Cloud's needs
- Video cloud technology stack
- Live video technology
- Some Thoughts
Foreword
Because some time ago to go out to do an exchange, so simple understanding of a moment in some cases related to cloud video industry, but also made some of their own understanding and problems. In this paper, with reference to the "2018 China Industry Research Report video cloud services" and content "AWS live video Program".
Traditional self-built video systems issue
The traditional self-built video system consists Cache servers, storage servers, server coding, scheduling server, other application servers, the bandwidth, IDC room, CDN node system maintenance personnel and developers. Since the video system with the dual demands of resources and technology, independent video system takes to build a very high cost, but in the complex environment of high concurrency, low latency network is still difficult to avoid Caton and other issues. Set up a video system first need to invest funds to purchase expensive infrastructure, and secondly, audio and video technology research and development process is complex, involving diverse technology stack, businesses in need of secondary development difficult and scenarios based on actual business needs, regardless of the human cost or time cost considerations are not a small investment. Due to exclusive resources and business performance of the entire server, but in practical applications rarely run at full capacity, resulting in a large number of long-term resource server is idle, widespread waste of resources.
High costs :
- Labor costs
- Time costs
- Construction costs
- Waste of resources costs
Poor results :
- High technical threshold
- Poor network environment
- User Experience Low
- AI is not energized
- No elastic expansion
Video Cloud (VaaS) the definition and origin
Defined : the concept of cloud-based computing technology of video streaming services, covering all the processes to capture video from video playback, allowing customers cost-efficient way to build a professional video systems.
Video Cloud provides a one-stop solution and a key to build the whole end of the broadcast service. Customers only need to use the video service cloud service vendors (usually in the form of SDK) to complete the set of processes the video player, as a user you do not control the entire video is how "coding - Distribution - Decode", you only need to It offers video content and video to the consumer. Video Cloud is highly market-oriented division of labor reflects greatly reduces video broadcast, on-demand, interactive threshold, also stimulated the live video of the wave of entrepreneurship and demand.
Video Cloud advantages :
- Low cost : a significant reduction in construction time and construction costs;
- High efficiency : can quickly clear statistical background video browsing information by visiting visitor's personal information to help customers better play the role of video;
- Elastic expansion : billing can be elastic, powerful statistical functions according to user specific circumstances;
- Globalization smooth playback of HD : video hosting multi-CDN distribution network acceleration technology, making video playback speed is greatly enhanced;
- Video Security : video hosting certain encryption technology uses such pirated video difficult to theft;
- Terminal capacity : you can artificially control the video ad insertion and whether the self-designed advertising, cloud services support smooth video, and other high-definition broadcast technology.
Video Cloud application scenarios
- Live matches (live)
- Video portal (VOD)
- Online education (interactive TV)
- Social application
- Remote Collaboration
- Security Monitoring
Demand and live video cloud is the most typical two service types. Cloud transcoding is the core technology on-demand cloud services through a server in the cloud video into standard formats, different bit rate based on the end user's bandwidth video files, to achieve both clear and smooth viewing experience. Live and on-demand more real-time and unexpected compared to cloud services and thus live more sensitive to delay and the first screen time and other indicators of stability under high concurrency and weak network environment also have higher requirements. And common way many different live, interactive live support for users of two-way communication. This two-way communication on the one hand, extremely low latency tolerance, on the other hand because it involves multiple video streams, codec bandwidth pressure and the pressure multiplied. Currently, video and more cloud vendors through a combination of real-time communication technologies to solve problems faced interactive live.
Video Cloud Market
The principle of cloud video
Video Cloud including cloud storage, cloud transcoding, live clouds, cloud encryption, cloud publishing and other functions. Video applications running on the cloud, the operation results (display output, audio output) coded through the network real-time transmission to the terminal, the terminal displays the real-time decoding outputs. Terminal and can be operated through the operation of the network control information transmitted in real time to the cloud application platform for running application control, the terminal can be streamlined to only provide network capability, video decoding capabilities and human-computer interaction.
Collection end : capture video content provider will first be encoded into a particular format; acquisition and rendering the interface will use the specific hardware platform, which interfaces the hardware and the specific design and performance are inseparable. Therefore, in the system design stage, we must consider compatibility and cross-platform hardware.
Server : the video after you've uploaded to the cloud server, usually in the cloud transcoding to adapt to different network environments and player terminal, the server will provide personalized value-added services, including statistical data, content auditing, real-time, including a watermark. Codec links would involve specific hardware processing power of the chip, there are two categories: hard and soft ed ed hardware solution software solutions.
- Live cloud, cloud-demand : Service capabilities
- Cloud storage : storage capacity (distributed cache, distributed storage), video data is stored, disaster / backup.
- Cloud transcoding, encryption Cloud : computing (CPU, GPU, FPGA), a video encoder, video rendering, high-definition image stream.
- Cloud released : CDN content distribution network.
Playout : Boost distributed via a content delivery network (CDN), ultimately played on the terminal device by decoding the content consumer.
- Player : interactive terminal (2D, 3D, VR)
Video Cloud Case
Universal Video Tencent cloud cloud scheme
Ali cloud video on demand solutions
Video Cloud's needs
Performance requirements
- High reliability : Anti-loss ratio exceeds 30%, the anti-800ms network jitter.
- High concurrency : ten million concurrent.
- Low latency : end of delay is less than 400ms.
- Security : encrypted transmission.
Functional Requirements
Live core requirements:
- Interactive features: support for text interaction barrage messages.
Audio and video applet core requirements:
- Enhance communication efficiency, enhance customer service experience: from the applet video calls, users face to face communication and customer service.
- Face to face counseling needs of the video: video service personnel know the user operations, solve problems, enhance the reputation of the manufacturer.
Education and training core needs:
- Personalized teaching: support for high-definition real-time video calls, real-time, QQ and anti-echo and other properties the same level.
- Immersive interactive experience: set audio and video calls, interactive whiteboard, IM, document transcoding, screen, and media sharing.
Short video core requirements:
- Fast uploads, real-time distribution: by quickly upload function, real-time sharing great content with others.
- Effects, features dynamic effect developed: shooting, filters beauty, pendant dynamic effect, effect plug-ins, green screen keys, editing and other functions.
- Social realization: short films and original film by switching the mask with a face pack, to achieve pay a reward function.
Video cloud technology stack
Coding
H.265 standard based on the H.264 standard, to improve the relationship between the code stream, encoding quality, delay and complexity of the algorithm, the standard can save about half the bandwidth at the same image quality.
1080P era, dominated by the H.264 video coding standard being shaken after the current video encoding has entered into. As early as 2013, a new generation of video coding standard H.265 to complete the development of standards, but is limited by the prevailing environmental conditions that are not widely popular. Compared to H.264, H.265's biggest advantage is basically unchanged under the premise of maintaining the image quality will be reduced to half the bandwidth of video transmission of H.264. The current mainstream video section or cloud vendors already support H.265 standard, or are planning to support the H.265 standard, along with future support for software and hardware environment, will become the standard H.265 standard video cloud services.
CDN + P2P technology
Cloud computing and CDN form complementary capabilities, significantly increasing storage capacity and resource utilization node CDN, and using dynamic node distribution technologies provide intelligent resource global perspective retrieval. P2P + CDN edge of the network through the use of fragmentation idle resources, improve content delivery network
transmission efficiency can be solved in seconds on video playback, delay and other problems.
In the streaming media distribution technology, P2P technology because full use of edge devices, thus easing the pain point of high-bandwidth video-cost cloud service, they have been using more and more video content distribution cloud vendors to the field. P2P scalability, high reliability CDN, combined with P2P CDN, the two complement each other, especially for the larger flow, heat and a high concentration of the same resource utilization scenarios.
Video codecs and content distribution networks are two key technologies in video cloud services.
cloud computing
** China's cloud vendors often use SaaS (software as a service) in conjunction with IaaS (infrastructure as a service), PaaS (Platform as a Service) and IaaS service model combined with the. ** Due to the network, storage and other infrastructure critical for audio and video services, the there IaaS service is indispensable. IaaS vendors will be more resources to basic extends the benefits of the business system, homeopathy provides PaaS cloud service levels. PaaS services in order to provide a common competency-based audio and video, all the needs of enterprises need to be developed based on business, this will make the company's products more flexible, more personalized, but also request for their own research and development capabilities. SaaS layer may provide a direct service available audio and video software, can also help enterprises to operation and maintenance. However, due to the high degree of standardization, coupled with the quite different demands of different customers, the SaaS service is more suitable for the development of the appropriate capabilities for specific vertical scenarios.
Close
PaaS layer to provide a universal service based video capabilities, usually for video-business-critical customer base, customers can be customized according to the specific needs of business development and optimization of the customer's own IT R & D capabilities have higher requirements.
PaaS vendors will be more focused in areas which there is a strong demand for rigidity video, which usually presents scenes of extending a general ability, presented in a modular fashion so that you can become more flexible to respond to customer needs. Pan-entertainment and online education PaaS vendors are generally optimistic about the two scenarios, where the pan-entertainment scene requires cloud services with a high degree sensitive to the changes in the market, even in hot pursuit can lead outlet, online education scene is required depending on the type of education and the audience to make customized adjustments.
SaaS
SaaS layer service is provided video software can be used directly compared PaaS, the IT R & D capabilities more suited to weak or non-critical video business customers. SaaS relatively standardized service layer, and between the demands of customers in different industries are quite different, SaaS vendors tend to develop more appropriate function for a specific vertical scenarios to meet the development needs of industrial customers depths with video services.
Because there is a big difference between PaaS and SaaS customer base faced, customers will choose different cloud service models based on demand characteristics, and thus relatively limited cross-vendor categories in the competition.
For medium and large enterprise customers traditional, standardized SaaS services are often unable to meet their needs, video and cloud need to get through the customer's own business systems, thus SaaS vendor service model will not completely limited to SaaS layer, will further border PaaS layer extends. In the relatively closed areas such as security and health care, more curable interests of the chain, and covers both hardware and software video services, both within the industry of video service providers have certain advantages to the cloud service transformation.
AI
Application of artificial intelligence technology to support online education, video surveillance, and other e-commerce scene to implement counter-terrorism Kam Wong, target detection and recognition, fragment detection, video deduplication and other functions.
The surge in the number of video content for each type of video platform in terms of both opportunities and challenges, in April 2018 ushered in the strongest short video industry regulation, vulgar, violence, gore, sex and other issues renewed attention. Simple manual review stage has been showing the way bottlenecks, combined with artificial intelligence video content audit labor costs due to significant savings in duplication of work brings, is becoming an important feature video cloud services.
In cloud computing, big data and artificial intelligence tend Trinity background, cloud vendors in video uploading, transcoding, distribution, playback on basic services, continuing to explore new value-added services, video content from the audit to directly drive cash flow and content producers to understand video content and video content editing. Although "AI +" level of penetration in the customer base is relatively limited, but taking into account the application of artificial intelligence technology will be able to bring efficiency gains, a significant role in value creation, thereby enhancing the value of video cloud services to customers, it is bound to become the next an important development direction of a stage industry.
Live video technology
Terminal Acquisition
Audio frequency
- The audio analog signal -> the PCM encoded original data -> compression coding -> package MP3, AAC, OGG, WMA, Opus, FLAC, APE, m4a, AMR and the like;
- The higher the sampling rate, the larger this audio signal recording data amount, the higher the quality of audio;
- Bit width: each sampling point needs to represent a value, the size of the data type may be 4bit, 8bit, 16bit, 32bit;
- Number of channels: audio capture and playback can be superimposed, the number of speakers corresponding to the number of sound source when the sound recording or playback;
- Audio frame: the audio data stream, there is no concept of a clear frame, generally the amount of 2.5ms ~ 60ms for a data unit of audio;
- The audio signal is assumed that a sampling rate of 8kHz, dual-channel, the bit width is 16bit, 20ms one, the size of an audio data:
size = 8000 x 2 x 16bit x 0.02s = 5120 bit = 640 byte.
Image :
- Cameras capturing apparatus into YUV coding original data -> compression coding H.264 / H.265 / VP8 / VP9 - > package MP4, 3GP, AVI,
MKV, WMV, MPG, the VOB, the FLV, the SWF, MOV, RMVB WebM and the like; - Image transmission format: common intermediate format (Common Intermediate Format);
- Image formats: YUV format usually stored in the original data information, wherein the image contains monochrome gradation values represented by 8 bits, and color images can be combined into three colors of RGB;
- Transmission channel: a video shooting normally only 1 channel;
- Resolution: video capture card can support a maximum lattice reflect the performance of its resolution;
- Sampling frequency: Sampling frequency reflects the speed and power for processing an image acquisition card. The higher the sampling rate, the higher the image quality, while preserving the data amount of the image information is greater;
Edge end treatment
Pre encoding compression processing, such as time-stamped, the company Logo watermark, beauty cream and confusion in the voice processing, which can be completed at the edge of the end or in the cloud.
Cloud codec
Video encoding significance :
- Original video data storage space, a 1080P video of 7s need 817MB;
- Original video data large transmission bandwidth, the bandwidth of 10Mbps transmission 7s video 11 minutes;
- After encoding H.264 compression, video size is only 708k, 10Mbps bandwidth only need 500ms, to meet the needs of real-time transmission, it is collected from sensors to capture video of the original video is bound to pass video encoding.
The basic principle :
- Spatial redundancy: there is a strong correlation between adjacent pixels image
- Temporal redundancy: content between adjacent video sequence similarity
- Coding redundancy: different probabilities of different pixel values occur
- Visual redundancy: the human visual system is not sensitive to certain details
- Knowledge redundancy: the regularity of the structure can be obtained prior knowledge and background knowledge
Coding parameters :
- Rate :
- Data transmission, data bits transmitted per unit time, is a key parameter to measure the size of the file size, a parameter indicating the number of KB per second, unit Kbps;
- The larger the rate is, the better the image quality, the greater the volume;
- Frames :
- FPS, the number of frames per second, i.e., how many pictures per second contained. The more the number of pictures contained per second, the picture appears more continuous, the less, the screen more "card."
- At the same bit rate, and inversely proportional to the image quality;
- 30fps generally acceptable, 60fps enhance the interactive sense and realism,> 75fps imperceptible;
- Screen size :
- Unit pixels, each 640x480 means that the video picture is 640 dots wide, 480 dots high.
- When a certain rate, and inversely proportional to the resolution;
- Mosaic : Levels rate is not caused by lack;
Plug flow
CDN transmission
Tree topology
- Backbone nodes
- Central node
- Regional node
- POP node
- Edge node
CDN-based tree topology, each layer has GSLB (Global Server Load Balancing) a plurality of load-balancing nodes CDN, Cache systems within the same layer.
Mesh topology
Mesh topology, each node is the Peer CDN, each serving peer node provides logically, not necessarily to complex design area network topology. The system can select any one of the fastest link through intelligent routing. Cache is not so high to previous requirements, the requirements of real-time is very high.
Live agreement
Some Thoughts
Private video cloud, public cloud hybrid architecture video
Common market video cloud solutions and more public cloud to provide, in the above video also repeatedly referred to the public cloud brings many benefits. But does this explain the necessary private video cloud did not exist yet? I think the answer is no.
In fact, as the business grows, users will be more willing to side through learning, while exploration, while the idea of constructing a set of self-control and gradually build private video cloud. Perhaps the beginning of dependence and learn to maintain posture to use public cloud video, but will eventually feeding into the private video cloud, combined with the hybrid cloud architecture to achieve two-way benefits. Of course, this approach to enterprise IT technology and capital strength have certain requirements, often seen in large enterprises. The key there is that the public - private issue of how responsibilities should be split between what the business on a private cloud, public cloud on what business, how to obtain the curve peak is the real value lies.
+ CDN edge technology brings the ultimate low latency
Low-latency video cloud certainly missed blessing CDN technology, but the essence is a CDN edge applications, is a subset of edge computing. Edge computing and CDN difference is: the former is considered force App and sinking, while the latter just a simple data caching and content distribution. With the maturity of edge computing is bound to the development of CDN technology has brought positive changes for the introduction of more CDN "software-defined" space.
Webcast cloud-based distributed applications. Created on the center of the cloud virtual machines run a streaming media server, on the edge of the cloud also create a virtual machine to run streaming media server, then using a mobile phone video streams to push the center of the cloud, you can see live video through mobile phone browser to access the edge of the cloud virtual machine and no matter how many users access the cloud edge streaming media server center to the edge of the video stream bandwidth consumed basically unchanged .
5G influence on the development of video cloud
5G arrival, the video industry will bring changes in technology. According to forecasts third-party organizations, 5G era, Internet video traffic per month from 2018 to 2022 the growth 100EB 270EB, which will result in the cost of the entire video distribution industry is increasing rapidly. However, there has been proposed a solution cloud vendors, Huawei cloud H.265 coding technique patent holder, to ensure at the same picture quality, the transcoding video bandwidth to achieve 10% to 20% lower than the average level in the industry, as well as customers realize lower distribution costs. Moreover, had had a live game show platform measured in a relatively simple scene live show live broadcast, Huawei's cloud transcoded 1080P screen compared to the average rate savings of 65%; in complex scenarios such as live gaming game , the average bit rate can save 41% . Visible, 5G make data easier to obtain, correspondingly, the data transfer bandwidth challenges. How narrow can narrow bandwidth (codec, encryption, compression) will be one of the core competitiveness of public video cloud, which is directly linked to the cost deduction.
5G edge technology and breaking the old technology framework for video cloud is no longer confined to a live, on-demand, interactive live scenarios. It inspired more imagination. Remotely operated excavators, telesurgery and other emerging application scenarios are endless. But because there is still a physical backbone network latency issues can not be avoided, 5G, audio and video data stream at the edge of the technical area blessing persists physical limitations. And this limit will stimulate the development of SD-WAN, leased line, VPN and other technologies, when cost backbone network leased line is greatly compressed, the real full-speed network era has arrived.