June 25th, The domestic well-known system high availability expert Series technology announced the open source of its core product capabilities, open to the outside world to produce the full link pressure measurement platform product source code, and officially named Takin.
At present, China Life Insurance, SF Express, Xyin, ZTO Express, China Mobile, Yonghui Supermarket, Aiinventory, Zhejiang University and other 50+ industry companies have adopted this technology, 100,000 + probes simultaneously online operation, real-time processing of tens of billions of user requests every day, generating hundreds of TB link data. The project has been released on Github. As the world’s first open source full-link pressure measurement platform, Takin’s open source will provide more enterprises with ultra-low threshold, ultra-low cost and ultra-high efficiency performance guarantee ability.
What is full link pressure measurement in production environment?
Full-link pressure testing is a method to achieve the most determined performance goals at the lowest cost, ensure business continuity, and enable IT systems to have anti-vulnerability capabilities and quickly locate problems.
IT systems is an engineer in combination with the specific business scenarios based on the basis of a series of components in coding, basic component itself limitations, as well as the uncertainty of the code, can make whole system exists a lot of uncertainty, this uncertainty will make the system in the face of a series of “risk” scenario (peak), being very vulnerable, So how do you make your system anti-vulnerable?
Pressure measurement through all production link, real simulation scenarios, “risk” business behavior real-time monitoring system, early recognition and rapid positioning system of the uncertain factors, and to deal with uncertain factors, optimizing the ratio of system resources, using minimum hardware cost, make the system easily face all kinds of “risk” scenario, achieve the expected performance of the system. In this way, a stable pressure measurement system can be implemented in the production environment to achieve long-term stable performance management of IT systems.
Performance testing has gone through four stages from offline to online evolution:
1. Demand-driven pressure test phase
Demand-driven pressure testing, most use simple tools for single interface or single system pressure testing, can also carry out some simple performance problems analysis, but most of the time there is no dedicated test team, need to develop independent pressure testing.
2. Performance regression system stage
Set up a special performance test team to build an offline performance test quality platform, and have the ability to perform full-link pressure measurement in complex scenarios and locate performance problems.
At this stage, there are three representative questions:
(1) Many companies have done offline performance tests, but there are still many problems online. The online environment is not well evaluated by the pressure test results of the test environment.
(2) With the growth of business and the increase of marketing activities, test engineers have no idea about activity security, and frequent problems in marketing activities will affect the company’s image.
(3) Performance pressure measurement efficiency can not meet the growing demand for performance pressure measurement, leading to some projects directly online without performance pressure measurement, and frequent online failures.
In order to solve the uncertainty of the performance pressure test environment, the performance pressure test began to evolve to the production environment, and entered the production environment performance pressure test stage.
3. Pressure test for production read-only services
In the stage of test environment regression system, the performance pressure test of read only business was added, and the production environment performance pressure test regression system was built to analyze the performance problems of read only business.
4. Full-service full-link pressure test
On the basis of the previous phase, the performance pressure test of write services is added to implement the full-link pressure test for all services. The performance pressure test and problem location capabilities of all services are provided. Better yet, system protection capabilities, such as degradation, current limiting, and fault drill, are also increased.
Why open source?
In an interview with InfoQ, Cao Xuefeng, CEO of Sequence Technology, said, “The original intention of Takin is simple: to enable more enterprises to use good products, to provide better user service experience, and to release more energy to expand their business. We believe that feedback on the use of the product itself has a positive effect on the development of the iteration, mutual benefit to achieve a virtuous cycle.”
At present, most enterprises still use the traditional performance pressure measurement method, but with the development of distributed, microservice architecture, this method can not meet the guarantee of system performance, Digitaltek decided to open source the production environment full link pressure measurement product and officially named Takin.
Of course, Takin does more than that. The biggest feature of open source is openness, inclusiveness and innovation. It is hoped that product open source can stimulate technological innovation in an open working way and attract more outstanding developers in the industry to join the co-creation team of full-link pressure measurement technology in production environment, so that the technology can be more landing and connect different use scenarios.
What is a Takin?
Takin is a production full-link pressure measurement system developed based on Java language. It can be embedded into each application node without the intrusion of business code to achieve full-link performance test in production environment. It is suitable for complex micro-service architecture system.
Takin has the following four characteristics:
(1) Intrusion of business code 0: there is no need to modify any business code when accessing, collecting and implementing logic control;
(2) Data security isolation: full-link performance test can be performed without polluting the business data in the production environment, and direct performance test can be performed on write interfaces in the production environment;
(3) Safety performance pressure test: performance pressure test in the production environment will not affect the business;
(4) Quick location of performance bottlenecks: The performance test results directly display the nodes with performance bottlenecks in service links.
What is Takin open sourcing?
Takin open source content mainly consists of three parts: Agent probe, console and big data module. The probe (Agent) is implanted in the Java application program, which can collect performance data, control the flow direction of test flow, and report the data to the big data module. The big data module will conduct some real-time calculation and analysis and store the data. The console is responsible for the management and presentation of these business processes. The three parts each play their respective roles to provide the business with non-code invasion, normal production environment full link pressure measurement services.
GitHub has the following open source address:
Takin:github.com/shulieTech/…
Open source community: news.shulie.io/? P =3024 (For detailed operation documentation)
There is still a long way to go in the future. Adhering to the original intention of helping enterprises solve microservice architecture governance and performance problems, Takin can greatly help enterprises reduce the development difficulty of producing full-link pressure measurement platform, and truly guarantee the performance and stability of more enterprise systems. We sincerely hope that Takin can join hands with the industry to build a more complete and standardized production full-link pressure measurement ecosystem.
Live preview: June 25, 7:30 PM, the two founders of zhonli nan and ping wei will answer questions for you in two live broadcast, interested partners can scan code into the group, now into the group will be reminded, there is iphone12 waiting for you to get oh!
