Tips: Estimated reading time: 3.1 minutes
Dear little friends:
Hello everyone, I'm Gavin, a 15-year IT veteran practitioners in the field of technology, work in BAT and other first-tier manufacturers nearly eight years, from CTO to start their own businesses.
Led the research and development of its own system, too explosive to hundreds of millions of high-volume platform has grown from several thousand subscribers, and upgrades along the way to face growing team of developers jumped after the amount of users and system technology systems With the development of more and more insights.
Looking back on the company's business has just started, the beginning of the C-terminal systems, management systems, databases on a single server, no operation and maintenance monitoring and release management for stable and healthy operation of the system can neither monitor nor acquisition system load conditions, for there is no user traffic is no major problem, but with the increase of users, complexity of back-end business reports to enhance the system to 502,503 errors occur frequently.
At first everyone thought system code is no problem, but there are also user feedback often point out the error to 502,503, while the current situation we have been numb. I personally stood in such a state, in-depth system architecture, rapid analysis locate the problem, first discovered all applications on one server deployments among businesses coupling is very high, which is typical of a single application, the system is very fragile, you need as quickly as possible decoupling and service transformation, I did the following eight measures:
1, the application server and database services and deployed individually isolated;
2, the database of the main service provided from the traffic separating applications to read and write;
3, will be a stateless application service transformation and distributed load balancing, support for flexible expansion;
4, application transformation services, infrastructure services sink, to do service isolation;
5, the data cache architecture and cache as a service;
6, the clustering settings and to ensure consistency with business services and service systems;
7, a key moment flow control, service degradation design, ensure that the main service stability;
8, online testing to ensure service stability and run-time monitoring of the service.
I was sort of our application framework of cross-border electricity supplier system, we can refer to
我通过上面的优化和改造后服务终于稳定下来并能支撑单日过亿的用户访问，服务上百个国家（我们目前的这套系统已经服务了来自全球140多个国家的用户，包含美国、中国、南非、加拿大、印度、菲律宾、墨西哥、澳大利亚、德国、西班牙、英国、秘鲁、以色列、法国、韩国、智利等国家。）
优化完成后，最终结果完全超出预期 ！
福利：下图是我整理的日均亿级项目实施敏捷开发流程规范 （首次分享）
随着平台用户量过亿后，项目开发人员也越来越多，各部门之间的业务交叉也变多，都有各自的代码标准，导致每次团队总结回顾会议进行代码review的时候都在讨论到底要按谁的标准来，都是各执一词没有统一的标准（互相不服，拍桌子），团队之间增加了大量的沟通成本，导致效率非常低下。
最终我们确定下来按照《阿里巴巴码出高效：Java开发手册》来进行代码规约，也因此很荣幸在2017年跟阿里巴巴码出高效Java开发手册作者孤尽老师认识并结缘至今；
今年9月阿里杭州云栖大会《大咖有约》现场
与孤尽老师再次深入交流探讨《软件工程危机与重构之道》
会后又跟孤尽老师也进行了深入的沟通：对于初创企业、中小型企业或者像阿里巴巴这样万亿级以上流量峰值的系统架构设计一定要谋定而后动，既要兼容过去的问题，解决当下的问题，还有适度解决未来的问题，而不是大而全大而广的进行系统架构。对于最新的系统的架构设计和重构方法我们达成高度共识。
孤尽老师亲自为我签名
《阿里巴巴码出高效：Java开发手册》20本
So, this will prepare me a chance to develop my experience of the past 15 years, one hundred million user platform architecture approach, project team management experience includes large flow such as Alibaba platform architecture design and other technology stack used to create a three days intensive training 2019 Ali Baba "one hundred million daily access to the system architecture design camp" this course full free .
1, attended the three days of training camp free courses;
2, Alibaba one hundred million daily access to the system architecture test.
You can receive free : Ali and I do solitary teacher personally signed "Ali Baba yards out efficient: the Java Development Manual" books one .
Scan the following QR code to add an assistant teacher micro-channel can be free to apply
First-served basis