On the morning of the sixth day of the sixth day, I just came back from my hometown and sat on the balcony of the rental house. The sun was shining brightly, and it turned out to be a rare good weather in Beijing. It has been a long time since the last time I wrote the year-end summary. I opened the blog and found that the last time I wrote the year-end summary was four years ago. The last time I wrote the summary, I was in the warm office of Dongzhimen. As I got older, I felt that time passed faster and faster. In four years, too many things happened: I had a child, changed jobs, Most importantly, it's thirty. Thirty means a lot of things. The ancients said that standing at thirty, for me, it is more important to have more responsibilities, not only family, but also work.
The first project I was in charge of at the beginning of the year was the operation and maintenance automation project of the configuration management group. Simply put, it was to use puppet to manage more than 20 machines that were manually managed before . Think about it, fate is really ironic. Just a year ago, in my last company, I was not too interested in continuous integration tools, and I was unwilling to learn them . puppet/chef , shell , I think these things are too trivial, and I just want to write code. When I changed jobs, I went to work in the configuration management team for a period of time. I had to learn these things. I won’t talk about the process. There is only one insight: many times, you think it’s too difficult, just because you don’t understand it . It took two weeks to build the entire puppet environment, everything is SVN , everything is code.
The next second project is to investigate and search for an automated release solution for the new architecture. This is a cross-departmental cooperation project spanning more than 20 project groups, including operation and maintenance colleagues, test colleagues and cloud computing. Colleagues from basic services, the research was conducted for a week. In fact, it took a long time to prepare in advance. The research plan for that week was revised to four editions. The structure of the whole new structure was systematically sorted out, and the leaders of the other party reached an agreement and obtained their support. , understand everyone's expectations: development colleagues hope to release code faster and more efficiently, test colleagues hope that the code to be tested is of the same origin as the released code, and operation and maintenance colleagues hope that the release process can comply with the specifications and controllable. When an agreement is reached, the plan is logical: the continuous integration server is responsible for one-click compiling, testing, packaging, uploading to the package server, and the package server saves all pre-release packages. One-click push to the Torca cluster scheduling system, Torca completes the release scheduling of the final cluster. Compared with the old architecture, I feel that the most obvious improvement of the new architecture is: the three modules of download, indexing and retrieval are separated into their own independent services and evolve independently; a unified data management platform, it was difficult to determine which module processes data when tracking bad cases If something goes wrong, now through the data management platform, the data processing process is visualized and traceable; a unified script execution system, all scripts and execution processes are transparently visualized; cloud computing platform, XFS file system, Xcube database, Torca cluster scheduling, mapreduce parallelism Computing, these services greatly simplify the development of upper-layer applications. More and more I realize the essence of the architecture: with the evolution of the system, we need to continuously decompose the system to achieve the independent evolution of services.Of course, there was confusion at the time: when all hope was placed on the new architecture, and all the success was accomplished in one battle, and the development and operation of the old architecture on the existing network ceased, the risks of the project can be imagined. After finishing this project, there are two insights: first, the opportunity only favors those who are prepared; second, the cross-departmental communication must find common interests, and must think more from the perspective.
In April , I am going to transfer back to the project management team and go to the cloud computing infrastructure department as a project manager. The last project in the configuration management group is Jenkins ' reporting system. It only took a week and a half. At first, I was going to use scala . Considering the subsequent maintenance, I finally used java . I haven't coded for a long time . key, build development environment, test environment and product environment, jenkins one-click automatic deployment, database version management, TDD , the first version will be launched in a week and a half, and finally I have to praise jenkins rest api . The insight is: Thank you that the development time of the first phase is only a week and a half, which makes us constantly think about what we need to do, which are the most urgent and which are the icing on the cake. After the first phase is launched, the only and the biggest benefit is: We no longer have to manually count and send weekly build reports, and we no longer have to be so busy every Monday. The time box is very important.
Finally, I turned back to the project manager and went to cloud computing, a place where great people gather. First of all, there are still supplementary courses: computer principles, Linux system programming, C++ primer , none of which can be missing. Not long after I went there, there was an accident: the search module's dependence on the cloud computing SDK is source code dependence. There are 5 products in cloud computing, but one product is incompatible with the previous SDK when it is released alone. The search module's failed to compile. I was responsible for promoting the solution of this problem, and set up a release process standardization project: by standardizing the release process and increasing automated integration tests, the release risk of the cloud computing platform was reduced. All SDKs are released on a unified basis. Automated integration tests must be performed before release, and compatibility tests with all SDK versions must be performed when the server is released. As the project progresses, it gradually integrates into this department: this is a department with a very strong engineer culture, everyone strives for excellence in technology, it is very common to work overtime after 10 o'clock, and the unit test coverage is amazing All reached 85% , but many colleagues have headaches when they hear the specifications and procedures. Project plans are also relatively random, and delays are common. In addition, because the previously released version upgrades are relatively casual, they will often receive complaints from upstream brothers and departments, and there are many occurrences. The problem, the brother department complained that the cloud computing platform was unstable, and after careful inspection, it was found that many times it was used in the wrong way, such as using traversal when searching for files. Gradually realize that the biggest problem of the department is the lack of product operation. Everyone's focus is on the product itself (throughput, maximum number of stored files, strong consistency), and users are more or less ignored. 5At the end of this month, the Fengshen project was launched. The goal of the project is to build a unified monitoring platform for typhoons and an automated deployment framework to create a one-stop typhoon service. Begin to introduce the practice of project management into the project. WBS is the most basic. The iteration plan finds the development rhythm, reviews, and strives to release the version online after each iteration to achieve continuous delivery. In engineering, the development environment and online The environment is isolated. The effect is not bad, but we should think more about what else we should do. To standardize product releases, it must pass automatic compatibility tests and inform users; modifications to the cluster environment must be auditable and cannot be automated for the time being, so colleagues in the department or paired operations must be informed first; Availability daily reports also need to be sent; development, testing, and live environments are isolated from each other; regular and proactive communication with users to understand their problems. The insight from this experience is simple: the core of a product lies in operation, and as a service department, we must deliver user satisfaction rather than products.
Immediately after that, the new structure has not been launched yet, and the organizational structure adjustment has come. I like the straightforwardness of ls : My current task is very simple, that is, I will level the mountain when I see it, everyone must listen to me, all People's thinking must be the same.
He delivered a speech "A Hundred Years of History to Look at Management" at the Agile China Conference. This session took 2 months of preparation and rethought the relationship between processes, organizational structures and people. From the beginning of the 20th century to the 1940s , management science completed the first stage of development from scratch. The most important achievement in this stage is to establish management as a science and discover the three elements of management: workflow, organization Structure and people, and tell everyone: management can be learned. We can see that the so-called management is just a balance and adjustment among the three processes, organizational structure and people. There is no fixed model for management, only different companies weigh and tailor these three according to different situations. If the first stage of management science was about how to do things right and how to improve work efficiency, then the second stage of management science in the 1950s and 1960s was about how to do the right things: Customer-centric, you must think clearly about the purpose of doing things before doing things. At this point, management finally has a complete definition: do the right thing, do the right thing. Since the 1970s , management science has entered the third stage of development. At this stage, the first idea proposed is that there is no silver bullet. Management is an art that requires flexibility. Next, the connotation of the process begins to extend, and it is no longer a simple work. The process is customer-oriented, emphasizing the entire process of meeting customer needs end-to-end. In the context of globalization, this process increasingly emphasizes the coordination between enterprises and the coordination of the entire delivery-oriented ecosystem. The concept of business process is proposed. Entering the new century, whether it is thinking about a more rational organizational structure (flattening), admiration for people (Jobs, innovation) or competition for business process efficiency (supply chain), they all tell us unmistakably: management only lasts forever Questions with no final answers .
Adjusted to a new department in September : Soso Ask. First responsible for the project management of the background group. The new team has only one old man, low morale, lack of documentation, hundreds of services, and it is very difficult to just maintain and rewrite the plan. Start with retrospective meetings and continue to improve. The insight during this period is: the best way to improve team morale is to help everyone succeed, and any achievement is worth encouraging . We have introduced continuous integration and automated release to encourage colleagues to summarize and share; automated testing has been introduced to encourage colleagues to report and help review ppt ; we actively let everyone be a programmer with attitude, think and give feedback on the product, and put the team together. The spirit is passed on to the department manager for encouragement. We can proudly say that the backstage team is now the most effective team. Another most important insight is that it must be the team leader who decides whether the team is competent or not. Fortunately, we have a very good leader .
In December , he was in charge of the department's community-based operation projects. This is consistent with the perception of this year's work. The core of the product lies in operation, which is exactly what I want to do. The project must have a name of NB , we call it the dark knight. This project also faces many challenges. At present, the biggest challenge lies in people. The confidence of the team has not yet been established. After the year, some people may propose to resign, and recruiting people is so difficult. Therefore, the second day of work The first thing is the retrospective meeting. The first version released by the team a year ago was very frustrating, the requirements were revised repeatedly, and the developers were discouraged. Therefore, the perception is: an excellent requirements document is the starting point for all reasonable plans .
In January , the departmental annual meeting program of the Technology Center was organized, and our original sketch "If You Ask, Do Not Disturb" won the second prize. Turning newcomers into protagonists is also part of team building.
Still thinking about it, what else should we do to ask. Has the traditional question-and-answer model as a supplement to search engines come to an end? Is the question and answer mode of SNS worth exploring? Are there deeper integrations with Weibo, or are they two presentations of one product in itself? Is Sina Micro's exploration not bold enough? On the mobile side, there is no prospect for an independent app , how can it be more powerfully integrated with WeChat?
Finally arrived at the paragraph that can end, and one thing seems to have forgotten to summarize, that is the book "The Eternal Way of Process - A Workflow and BPM Project in Action" that we wrote for four years, and said nothing Here's an example to illustrate why it's worth looking forward to: after we have connected all the data and processes of the Housing Administration Bureau and various commissions and bureaus with BPM , the customer still insists on manually stamping the seal and going through the manual process. The BPM implementation technology is not a bottleneck at all. The bottleneck is still people. It will be published in the first half of this year. The reason why I wrote it for four years is because as I write, I always feel that I don’t know enough, and I keep reading and supplementing the content. Really, I was young at that time, and the ignorant were fearless, alas.
2013 , The Dark Knight Rises!