High availability, full integration, customization, ant Financial automated testing evolution

The little ant says:

On November 16, pei Yang, a development engineer of Ant Financial Engineering technology Department, summarized and shared the development and practice of wireless experimental cluster in Ant Financial at the first Android Green Alliance developer conference.

There have been many practices throughout the industry regarding the application of automated testing in specific business scenarios. As DevOps, automated testing, and manual testing converge, a variety of development frameworks and design concepts are frequently created. How to support rapid business iteration, how to face the new technical solutions of dynamic and componentized clients, how to support automated testing of new IOT, human-computer interaction and other business scenarios, and how to conduct in-depth thinking and design of test cases under business logic have become the focus.

How did Ant Financial build an automated testing architecture through wireless experimental clusters in an application with 870 million annual active users? What specific optimization plans did ant Financial accumulate during this period?

High availability is necessary to support the business.

Mobile cloud testing platform as a necessary infrastructure to support automated testing, now in various manufacturers have the corresponding implementation scheme. At present, ant’s internal cloud measurement platform needs to have high availability and efficient and flexible execution and control capabilities. Only flexible and efficient high availability service capability can effectively improve r&d efficiency and reduce R&D cost.

1.1 Device Cluster

One of the main differences between mobile terminals and traditional servers is that there is a big gap in device reliability. High availability service SLAs can be guaranteed by a certain degree of redundancy combined with reasonable scheduling policies. Ants internal cloud service SLA requirements for key business is need reached the standard of 4, 9, and this target is based on equipment is two-way choice competition mechanism with real-time anomaly detection and dynamic task switching mechanism, because automation task granularity than traditional infrastructure services (Nginx, MySQL, etc.) the request of the particle size, Therefore, some special scheduling policies are needed to ensure.

1.2 Data Analysis

In order to ensure high availability, on the one hand, it is necessary to uniformly monitor all kinds of global resources, on the other hand, it is necessary to continuously analyze and mine buried data, and make continuous improvement for the problems found. Ant cloud testing service establishes hundreds of different types of buried point data reports for equipment and task execution, covering equipment health, task execution efficiency, time consuming of key scenes and other multi-angle data. In the process of automatic task execution, buried point analysis data of business dimension can also be dynamically added through standard extended interface.

1.3 Continuous improvement

Through continuous analysis of the data, we found a number of issues, including many difficult NREP issues, which were quickly fixed with the call stack records. Up to now, Ant Cloud testing has been iterating at least one version per week. It can be said that cloud testing services completely follow the model of agile development in small steps from the first day, which is an important guarantee that the platform can always support rapid business iteration.

How services can be better combined with business landing, is the best standard to evaluate the value of services.

2.1 Service Standardization

Ant experimental cluster in the actual application scenario, fully consider how to combine with business landing strategy. From the point of service, provide standard internal region of the RPC service interface, for other network isolation business domain provides a safe and reliable standard service interface, through a unified gateway service layer, on the one hand, access to business very simple, on the other hand, can according to different business scenarios and requirements of the corresponding business adequately isolation and current limiting. On the premise that service requirements are fully met, abnormal problems can be controlled and isolated at any time. Currently, the platform services receive more than 200,000 requests per day.

2.2 Service individuation

From the perspective of service content, Ant Cloud testing provides standardized technical solutions and entrances for general business needs; Business demands for special scenarios allow users to quickly assemble personalized task construction modes. At present, more than 200 special services of business lines have been running on the cloud testing platform for a long time, meeting the automation requirements of more than 40 applications development.

Another aspect of service personalization is that it can support more and more new business scenarios, such as human-computer interaction (face recognition), IoT (scanning code, screen casting) and so on. Traditional automation solutions focus on the terminal device itself and generally cannot include the physical scene of verification, while new business scenarios also have high requirements on the external environment that the device depends on. In this regard, the engineers of Ant Experimental Platform technology make full use of the combination of software and hardware to build various complex environments required by business through a large number of self-developed MCU devices and high-precision industrial robots. For example, the whole process simulation of scanning code for users entering the store includes possible lighting conditions. The displacement, acceleration, Angle, and free combination of terminal devices as the user pulls out the device.

Through the construction of these automated test scenarios, Alipay Wallet can continuously optimize the user experience.

Reduce operation and maintenance costs and improve r&d performance through high availability services.

With the continuous expansion of service scope of experimental cluster, the requirements and investment cost of resource management operation and maintenance are becoming higher and higher. In order to control operation and maintenance costs and ensure service reliability, ant Experimental Platform technology students designed an integrated solution based on AIO intelligent cabinet one-stop.

3.1 Flexible capacity expansion and on-demand combination

Each AIO cabinet is a miniaturized cluster of devices that provides the following capabilities:

• Custom MCU, packaging 9 types of 67 custom control instructions, covering data, voltage, current, temperature, power consumption control

• Controlled execution environment (light, electromagnetic shielding)

• High-precision power acquisition measurement capability (±0.01 coulomb)

• Simulation of various links (WLAN/4G/ weak network)

• Equipment special state simulation and protection (low power maintenance/charge/discharge protection)

These control capabilities can be efficiently integrated into a smart cabinet, or a subset of these capabilities can be adopted based on business requirements, with proven capabilities in cost control and product customization.

3.2 Service plug-in and hardware modularization

AIO cabinets have corresponding host services, and service capabilities are combined in plug-in mode. If users have special types of devices or special business requirements, they can quickly define new service plug-ins according to plug-in specifications to expand AIO capabilities. At the same time, AIO core main control board is also fully modular design, rich extension OF GPIO can facilitate the expansion of new modular hardware capabilities, to meet the demands of complex IoT scenarios in the future.

Up to now, AIO wireless experimental cluster has completed 50W + automation tasks in alipay system, executed more than 400W use cases and captured flashback 5W + times.

In terms of reducing the cost of manual testing and helping the ability of manual testing to expand to more application scenarios, we have successfully supported the test tasks of 40+ Ali Group App, inspected and found 250W + abnormal items, completed 15W + small program access review, and remotely shared equipment reached 4000 hours.

AIO wireless experimental cluster has now become a part of Ant Financial mPaaS. With the external output capacity of “MTP mobile Test platform”, it provides effective test schemes covering all stages of App development, testing and launching, effectively helping to save resource input in product testing and improve test efficiency and quality.