The following content is from “2021 Regional Distributed Cloud Conference: Theme Sharing of Serverless Cloud Native Unleash The Efficiency Dividend of Cloud Computing”. The in-depth article is expected to take 22 minutes to read.
Share the guest
Li Xiaochuan, an expert of Tencent cloud Serverless technology products, is currently responsible for the design of Tencent cloud Serverless products and enterprise solutions, committed to promoting the evolution and implementation of cloud native technology architecture, and providing efficient cloud service products for technical workers. With more than ten years of experience in technology research and development, he once worked in Oracle and IBM, engaged in the research and development of enterprise-level business suite and cloud service engine. Then he switched to consumer Internet, incubated MIOT and established MIOT e-commerce business. Before joining Tencent Cloud, he served as the technical director of IQiyi and the technical director of business line.
01. What does the business-oriented technology system focus on?
First, take a business perspective on why you need to focus on Serverless. The business perspective here is broad business, meeting needs, addressing pain points, and delivering value. The enablement of technology to business can be summarized in three stages:
- Iterate on the prototype and validate the business
From zero to one, improve the efficiency of iteration and validation of business prototypes through technical means;
- Continuous optimization, strengthen the business
Continuously optimize operational efficiency and engineering efficiency to enhance business competitiveness;
- Build barriers to enable business
Form a moat in the depth of technology, and continue to maintain the differentiation advantage of the business;
In order to better support the business, in the technical system we will focus on what issues?
**1. Stability and quality, ** is the cornerstone of normal business promotion;
**2. Development efficiency: ** helps the business gain precious time, which determines whether the business can win on the track, such as whether a new idea can be verified quickly;
3. Operation efficiency is the eternal persistent pursuit of r&d students. More optimized algorithms and more elegant architectures will also determine the depth and difficulty;
** The cost here, on the one hand, is the cost of resources, such as the cost of computing, network, storage and other basic resources needed to run the service or system. On the other hand, time and human cost, for enterprises, especially Internet or business-oriented companies, human cost accounts for a large proportion of the overall enterprise;
02. What dividends can Serverless release?
From the perspective of industry, there are four dimensions:
- Degree of modularity
- Degree of automated operation and maintenance
- The efficiency of the elastic
- Failure recovery capability
It can be seen from these aspects that the judgment of trend evolution of infrastructure layer has been made. At present, Serverless is an inevitable evolution direction of computing platform in the foreseeable stage, and also the final direction that can be foreseen at present.
Serverless is divided into two stages: the first stage is containerized Serverless, and the next stage is the platform for fPaaS Serverless.
As for the definition of Serverless in the academic world, University of California, Berkeley published a paper in 2009, which accurately predicted the evolution direction of cloud computing in the next 10 years. In 2019, another paper was published, predicting that Serverless would become the final product form or service platform of cloud computing in the future.
Tencent Cloud held ServerlessDays China Communication Summit for two consecutive years. In 2020, Johann Schleier-Smith, one of the main authors of this article, was invited. What impressed me most was that he divided cloud computing into two stages. Simplify system management, including VMS and containers, and provide a programmable and simplified management mode. The cloud platform hosts the infrastructure. The next stage: simplify development, developers only focus on the core code, cloud platform is responsible for system management, operation and maintenance, architecture, etc., to provide better experience for development, so as to improve development efficiency.
After explaining industry and academic definitions, let’s compare Serverless with traditional Server.
In the traditional Server pyramid, the security and scheduling of system resources at the bottom level, logging, monitoring, operation and maintenance at the middle level, data and storage at the top level, and finally the business logic at the top of the pyramid. For business development, what really generates value is the writing of business logic at the top of the pyramid, serving customers and bringing value to customers. The following part is more basic support. From this point of view, Serverless is to solve the following part of infrastructure maintenance. The platform side undertakes the bottom part of the foundation, so as to give developers Less operation and maintenance costs and workload.
Taking business development as an example, when planning a project, about 50% of the time is spent designing the entire architecture, such as high availability, high concurrency and scalability. How to design the architecture to support such a scale? If the platform side undertakes this part of the work, it can liberate a lot of manpower and energy.
For the change of business model, in some typical scenes, such as “e-commerce killing in seconds” and “Snatching red envelopes in the Spring Festival Gala”, there are instantaneous peak traffic. In order to undertake such instantaneous peak traffic, huge resources need to be reserved, but these resources are idle for a long period of time.
How to solve the problem of resource redundancy?
This business model has been implemented in all sectors of clothing, food, housing and transportation in the past, but it is reused in the field of computing. In the early stage, we built our own computer room, then started to rent cloud host, and finally formed pay-as-you-go through Serverless. Analog traffic, buy a car, pay for maintenance, depreciation costs and take up their own time to drive; If you rent a car, you need to pay the rent if you use your own time to drive or not; Taxi, on demand, mileage charge.
The cloud vendor has the right to use the computing resources provided by the cloud vendor during the lease period. However, some computing resources are redundant during the lease process. The Serverless pays on demand applies for the use of computing services based on the actual demand.
From this perspective, the core values of Serverless can be summarized into three points:
1. Simplifies operation and maintenance, provides a stable operating environment, supports multiple scenarios and elastic scaling services;
2. Lower cost, pay on demand, reduce the billing of redundant resources;
3. Time efficiency: Serverless, an event-oriented programming mode, greatly simplifies architecture design and team collaboration with different specifications and standards. Fast business landing, quickly put into the market to do verification.
03. Serverless landing mode
1. Tencent Cloud Serverless product matrix
- Serverless HTTP
Currently support wechat small program development platform, more typical Web applications, API services and so on. Cloud functions have supported more than 1 million developers on the wechat development platform, with more than 15 billion calls per day.
- Serverless Trigger
Triggers are “events” defined by components on the cloud that trigger calls to functions through an event-driven model.
The above two points is a lightweight calls and services, audio and video processing and data processing ETL belong to heavy computing scenario, further subdivided and understandable for audio and video processing is for the processing of unstructured data, including the transcoding, understanding, data processing, ETL is the processing of structured data, including data analysis, summary, calculation.
The above is more about the product capabilities of Serverless from the perspective of application. Serverless Framework and cloud function SCF are the basic product support, and cloud function is a new generation of computing platform that supports all the computing requirements of the upper layer applications. Serverless Framework is a cooperative product between Tencent Cloud and Serverless.com in North America. It is the most popular Serverless development tool in North America.
2. Application scenarios of Tencent Cloud Serverless product capabilities
- Serverless HTTP product capabilities
In the traditional service mode, you need to create compute instances for Web applications and API servers, deploy application servers on compute instances, and then deploy release codes. After that, you can bind public network portals to connect to services to form a service closed loop. If it is realized through Serverless, API gateway and cloud functions can be naturally integrated. You only need to write business logic in the function. If the logic is clear, development, debugging and release can be completed in a few minutes, which greatly improves the efficiency of business iteration.
- Serverless trigger
Based on the cloud native architecture, various components on the cloud, such as COS object storage, Ckafka message queue middleware, CLS log service and so on, when the defined event is triggered, the built-in trigger of the component will call the cloud function to process the corresponding business logic. The product capabilities of COS object storage triggers and functions such as long running and large size instances form the basis of Serverless multimedia solutions.
- Serverless Audio and video processing
For developers, we will build mid-Taiwan services, mainly providing three differentiated capabilities:
-
The calculation force can be configured, and the calculation instance size can be explicitly selected through the function specification.
-
The framework/encoder can be configured to support deployment of self-developed or open source encoders to achieve smooth migration of original transcoding or audio and video processing services.
-
Workflow can be configured, and the upstream and downstream industrial chains and business processes can be connected in series to realize the closed loop of the whole business chain.
On the basis of transcoding, Serverless cloud function provides more in-depth capabilities, and has jointly explored many solutions, including intelligent audit, image quality enhancement, video understanding, video editing and so on, with the deep accumulation of laboratories within Tencent Cloud.
For real-time audio and video interaction, Serverless cloud function also provides encapsulation of such capabilities as ** one-click trigger live push, post-processing of by-pass recording of live stream, review of content, enhancement of original content (real-time captioning, real-time translation) and so on.
- Serverless data ETL
In addition to some processing of unstructured data, structured data ETL processing is also supported. By pulling multiple data sources such as public network API, DB and file storage, batch processing and loading of large amounts of data to target storage, scenarios such as e-commerce, search, content, monitoring and analysis are covered.
Finally, I will share with you the existing enterprise-level solutions to help more enterprises achieve business goals efficiently:
- Serverless Multimedia processing
- Serverless Data processing
- Serveless AI reasoning
- Serverless HTTP
That’s my share. Thank you.
Author: Li Xiaochuan.
One More Thing
Experience Tencent Cloud Serverless Demo immediately and receive Serverless new user package 👉 Tencent Cloud Serverless novice experience
Welcome to: Serverless Chinese!