Shulou(Shulou.com)06/03 Report--

At present, Serverless has become one of the focuses of the cloud native community, some people say that it is the successor of micro-services, will completely change the status quo of software development, so what is the real situation? This article will introduce the Serverless market observation, landing challenges, as well as Ant Financial Services Group's practice of Serverless.

Serverless has become a trend

When we review the development of cloud computing, we will see that the infrastructure has undergone an evolution from physical machines to virtual machines, from virtual machines to containers. Under this general trend, the application architecture is also evolving synchronously, from single to multi-tier, and then to the current micro-services. Behind the change, there is a continuous driving force, which comes from three constant pursuits: improving resource utilization, optimizing the development, operation and maintenance experience, and better supporting business development.

At present, Serverless has become one of the focuses of the cloud native community, and its development is no exception. Compared with container technology, Serverless can refine the granularity of resource management, enable developers to get started with cloud natives more quickly, and advocate event-driven models to support business development. Thus, it helps users to solve the problems such as the complexity of resource management and the occupation of low-frequency services, and realizes resource-oriented instead of resource-oriented allocation mode. According to a CNCF report based on 2400 people at the end of 2018, 38 per cent of organizations are already using Serverless technology, an increase of 22 per cent over the same period in 2017. (source: CNCF Survey)

At present, cloud vendors offer a variety of Serverless products and solutions in the market, which can be roughly divided into:

1. Function computing services, such as AWS Lambda, are characterized by running in snippets of code and have certain requirements for code style.

two。 Application-oriented Serverless services, such as Knative, are based on container services and provide build capabilities from code packages to images.

3. Container hosting service: such as AWS Fargate, which runs in a container image unit, but users still need to perceive the container.

From a community perspective, the CNCF Cloud Native Foundation is coordinating community discussions and promoting the formation of specifications and standards through the Serverless working group, which has produced important content such as Serverless white papers and panoramas. Among them, the panorama divides the current ecology into four modules: platform layer, frame layer, tool chain layer and security layer.

Photo Source: https://landscape.cncf.io/

Landing challenge

In the process of communication, we found that Serverless well addressed some of the customers' demands, including increasing resource usage and reducing costs through 0-1-0 scalability; supporting code packages to enable customers to realize cloud native without impact, and historical applications without container transformation; supporting flexible trigger configuration to guide users to carry out event-driven transformation of applications, so as to adapt to the rapid development of business, and so on. These advantages make Serverless shine brilliantly in the context of Mini Program development. At the same time, there are many explorations and breakthroughs for Serverless landing in the production environment of enterprise applications. At the KubeConChina 2019 conference some time ago, the Serverless working group meeting also discussed this topic. The current core challenges can be summarized as follows:

1. Platform can be migrated

At present, many platforms have launched their own Serverless standards, including code format, framework and operation and maintenance tools. Users not only face high learning costs and choice pressure, but also worry about being unable to flexibly migrate Serverless applications between platforms.

2.0-M-N performance

Online applications have strict requirements to control request delay, so users need to carefully verify that the Serverless 0-1 cold start speed, MmurN expansion speed and stability meet the production requirements.

3. Debugging and monitoring

Users are not aware of the underlying resources and can only debug and monitor the application with the help of the platform capabilities. Users need the platform to provide powerful log functions for troubleshooting and multi-dimensional monitoring functions to keep abreast of the application status.

4. Event source integration

With the Serverless architecture, applications tend to be split in finer granularity and concatenated through events. Therefore, users hope that the platform can integrate most common event sources and support custom events, making the trigger mechanism more flexible.

5. Workflow support

The completion of a business often involves the cooperation between multiple Serverless applications. When there are a large number of applications, users hope to use workflow tools for unified orchestration and status checking to improve efficiency.

Ant Financial Services Group's practice

SOFAStack is committed to solving the actual pain points of customers on the cloud through product technology, precipitating Ant Financial Services Group's technical practice, and helping users migrate to cloud native architecture in an efficient and low-cost way. Serverless Application Services (Serverless Application Service, referred to as SOFA SAS) is an one-stop Serverless platform derived from Ant Financial Services Group's practice. SAS is based on SOFAStack CAFE Cloud Application engine (Cloud ApplicationFabric Engine referred to as CAFE). CAFE's container service has been certified by CNCF and is a standard Kubernetes.

While being compatible with the standard Knative, Serverless application service products incorporate the application lifecycle management capabilities derived from Ant Financial Services Group's practice, providing supporting capabilities such as Serverless engine management, application and service management, version management and flow control, 0-M-N-0 automatic scaling, metering, logging and monitoring, which can be triggered quickly according to business requests or events. At the same time, combined with the actual pain points of customers on the financial cloud, the product ingenuously provides two forms of exclusive version and shared version, as well as three R & D modes of traditional code package, container image and pure function, in order to solve the different needs of users and reduce the barriers to customer entry.

One-click deployment: users can deploy applications with one click through code packages or container images and test and execute them at any time.

Engine management: SAS provides rich engine lifecycle management, diagnosis, monitoring and other capabilities to provide exclusive customers with all-round management and operation capabilities of Serverless engine data.

Services and versions: SAS provides application management, application service management and version management. The version can be deployed as a container image or as a code package under the traditional VM release mode. In many cases, user code can be migrated without modification or writing maintenance Dockerfile.

0-M-N:SAS provides the Serverelss rapid scaling capability of 0-M-N-M-0, supporting event-triggered or traffic-triggered 0murm, and Mmurn of multiple metrics (such as QPS, CPU, MEM, etc.)

Log monitoring and measurement: the product has built-in log, monitoring, metering and other supporting facilities to help users debug and monitor the application status.

Flow control: basic flow control capabilities are provided based on SOFAMesh,SAS, which will be further integrated with service grid to provide large-scale, multi-dimensional, cross-regional and hybrid cloud flow control capabilities.

Trigger management: the product supports cron expression triggers based on common cycles and second precision, which can correlate and trigger serverless applications, and will support more IaaS, PaaS managed and data events in the future.

Performance analysis: the horizontal axis is the number of Java applications that completely trigger cold start at the same time, and the vertical axis is the average and minimum time-consuming of cold start applications. With the increase of pressure, it takes about 2.5 seconds for 50 Java applications to schedule plus cold start at the same time, and 3-4 seconds for 100 Java applications to schedule cold start at the same time, with a minimum time of 1.5 to 2 seconds.

Performance brief Analysis: Pooling fast bullet slow shrinking timing algorithm, the relationship between pool capacity and actual unit time applications can be achieved as shown in the figure (blue is the actual number of applications, green is the pool capacity). At present, the product has successfully supported the production environment Mini Program Serverless model. At the same time, the ability of 0-M-N-M-0 has greatly reduced the operating costs of Mini Program. In the field of industry customers, an insurance company decided to move some of the day-end front and long tail to the Serverless product platform in the near future, which is another important breakthrough for our products. In the future, we are committed to building SAS into a financial Serverless platform.

Serverless application service was officially unveiled at KubeCon China 2019, attracting more than 100 participants in the SOFAStack workshop to experience the easy construction of cloud applications based on Serverless.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.