Skip to main content
Background Image
  1. Miscs/

Baiji 1509 Training Reflections

·1813 words·9 mins· ·
Ruohang Feng
Author
Ruohang Feng
Pigsty Founder, @Vonng

On September 23rd, I came to Hangzhou to attend Baiji training. Unlike Bai’a, I didn’t have special expectations for Baiji beforehand. After all, for a one-day training, it might just be going through the motions, and I was skeptical about how much I could learn. However, after attending a full day of classes today, I think Baiji was indeed worthwhile. Striking while the iron is hot to submit my homework, I’m writing this reflection.

Now, when it’s fresh —— Dr.Grace, Avatar

Today there were five main speakers: Yisu, Shen Xun, Fan Yu, Nantian, and Xuannan.

In my view, Teacher Xuannan’s class was most inspiring and helpful, Teacher Nantian’s class was full of insights and humorous, Teacher Shen Xun’s presentation was the most vivid and easy to understand, Teacher Yisu was promoting their DingTalk, while Teacher Fan Yu’s class had the most senior colleague vibe.

First up was Yisu, senior architect from the DingTalk team. The first lecture was called “Past and Present Series - The History of Ding.” DingTalk’s advertising is indeed loud - there are DingTalk ads in the elevator where I’m staying, so for the first question of the first class about DingTalk’s slogan, I got first blood and won a Taobao open source badge. Teacher Yisu explained DingTalk’s design philosophy: the concept of “work circle” corresponding to “friend circle.” He also demonstrated the “Ding” function and conference call features on site. Practical skills, practical tools. However, in my opinion, this might not be very suitable as an opening lecture for Baiji, since “what is above form is called the Way, what is below form is called instruments.” At the end of the first class, Teacher Yisu raised questions like “how to optimize connections in weak network environments, how to optimize image and voice transmission” to inspire everyone’s thinking. In my view, this should have been the most valuable part of this class, but unfortunately it was cut short due to time constraints. It was quite a pity.

The second class, “Past and Present Series - The Rocky Road of TAOBAO.COM,” was taught by Shen Xun, senior technical expert in middleware. This class was very well taught, following the historical development sequence with very clear thinking. He introduced the evolution of technical architecture during Taobao’s development process and the many problems faced. Bandwidth pressure from Taobao traffic growth, database load increase problems, development costs skyrocketing with team scale, hardware relay bottleneck problems, etc. Since many of these things were done by him personally, his explanations hit the essence of thought directly. Many of these problem solutions are familiar, and their solution ideas are all very simple and intuitive. In a nutshell: one universal pill - middleware decoupling; two optimization axes - Parallel & Hierarchy. Of course, I’m also very clear that simple solution ideas definitely don’t mean simple implementation difficulty, as there are too many corner cases to solve.

I won’t elaborate on the lecture content details, but what made me happy was Teacher Shen Xun’s teaching method, which I must specifically mention. Teacher Shen’s teaching method had two characteristics: first, using a history-oriented teaching approach; second, focusing on cognition and insight rather than knowledge transmission, emphasizing technical thinking rather than implementation details. (Actually, Teachers Nantian and Xuannan both did this as well, or even better)

From Taobao’s initial (WebApp—>DB) architecture to today’s seemingly dazzling complex system, for every architectural upgrade, Teacher Shen Xun would point out what the problem was and where it came from. He told us the ideas behind solutions rather than details, and finally the application effects. Many courses and training sessions like to extensively explain “ingenious technical implementation details.” But what I care about is what problems existed originally, how these problems were discovered, what the solution approach was, and what effects were finally achieved. For those domain-specific tricks I’ll probably never use in my lifetime, none give a shit. Those are knowledge indeed, but what’s truly precious are: first, the cognition to keenly see problems and think of solution approaches; second, the insight to use knowledge to solve problems. For each stage of Taobao’s development, Teacher Shen explained his views on problems and his thinking process, which are undoubtedly precious.

Another characteristic, precisely related to the lecture series name “Past and Present,” is the history-oriented learning method. I’ve consistently used this method since I encountered philosophy, as Hegel said: philosophy is the history of philosophy. I believe that in many disciplines, the position of disciplinary history is greatly underestimated, though this trend is slightly weaker in project practice. In some textbooks and project documentation, there are often so-called “knowledge crystals.” Crystals are beautiful, but we cannot understand the environment and conditions when crystals formed, or discover the mechanisms behind crystal formation, just by observing their form at this moment. Understanding knowledge itself means mastering the current state, while understanding history means grasping development trends. Only when position and velocity are simultaneously determined can we make accurate judgments about the future and make correct decisions. Without understanding and learning from history, we cannot understand the source of problems, fundamental needs, and the historical limitations of solutions in that specific period. Knowledge and methods only constitute complete solutions together with their supporting applicable environments. I’m glad Teacher Shen’s lecture didn’t just take out Taobao’s current architecture, break it into components and blah blah about them, but instead used a holistic approach and evolutionary form to explain its evolutionary process, which felt very effective. This lecture series truly lives up to the “Past and Present” series title.

Let me insert a digression here - knowledge graphs were also mentioned in the lecture. I also have some thoughts on this, which happen to be related to the history-oriented method. Many current “knowledge graphs” have a problem - knowledge graphs are network structures, but they should definitely not be flat two-dimensional networks. Their third dimension is the time dimension, the historical dimension. Connections in knowledge graphs should include not only original connections between concepts within each time slice plane, but more importantly, associations across time slices. Limited by two-dimensional presentation forms, maybe this can only be an idea, but perhaps recent virtual reality and augmented reality technologies can bring new opportunities for presenting such knowledge graphs.

Finally, because I happened to be working on crawlers and recommendation systems during my internship, Teacher Shen’s first and only question in the lecture - “describe how search engines work in five sentences” - let me get another first blood. Two prize tickets, haha.

The fourth course was “Past and Present Series - These Years of Double 11.” The speaker was Nantian, senior director responsible for Tmall Double 11.

If Teacher Shen’s lecture deserves 9 points, then Teacher Nantian’s lecture deserves 10 points. The Double 11 story was told very well! I’ll have bragging rights in the future (laughs :D). The main reason is: if Teacher Shen’s discussion of Taobao’s entire technical transformation might seem somewhat distant from us, then Teacher Nantian’s specific project that everyone is familiar with and personally participated in (as users) is obviously very down-to-earth. High energy throughout, besides various first-hand exciting stories, this lecture’s insights were divided into two parts: viewing projects from a leader’s perspective in all aspects, and the technical transformation process driven by specific needs.

Also about technical architecture evolution, having just attended Teacher Shen’s training, Teacher Nantian’s lecture seemed very accessible. In this lecture, aside from knowledge content about Tmall Double 11 and understanding of project development and transitions, two points left deep impressions on me.

The first point is about two core problems of technical limitations: scalability and development conflicts. These two problems have been clearly discussed in Harvard E-75 and “The Mythical Man-Month” respectively. I’d rather understand these two problems as: money can’t solve problems anymore and people can’t solve problems anymore. These two problems always appear alternately in the race between technical level and business needs. If the last time cloud migration solved the scalability crisis, then the next business growth bottleneck might very likely occur again in technical coordination, and some signs can already be seen. (Rather than treating coordination development crises as technical problems, it’s really better to say they’re cost problems of management and communication)

The second point is about the advancement mode of technical progress: evolution and planning. This is somewhat similar to the difference between neural networks and expert systems. I asked Teacher Nantian about his views on future technical bottlenecks. He believes our technology is evolutionary, driven by needs. Who knows where the future will go? I won’t elaborate here. Anyway, this lecture was also very powerful.

Finally, the concluding lecture “Technology Shapes Life” by researcher Teacher Xuannan, I think was the most exciting course in the entire Baiji training. What I learned in this class was invaluable - truly precious life experience. (Taking a moment to continue). Basically, all the viewpoints Teacher Xuannan raised formed a superset of my views on these issues, so I felt truly excited while listening, filled with regret.

The outline is roughly as follows:

  • Programmers’ vision problems: triggering thoughts about breadth vs. depth. Top talent’s skill composition should be one specialization with multiple strengths - professional expertise + universal interface. As programmers, we should actively understand business and think about products (but don’t overstep and do PM’s work). This could be another article.
  • Development model issues: the meaning behind agile development and rapid iteration. I’ve gradually shifted from efficiency obsession to pursuing flexibility and adaptability as the highest goal.
  • Focus on human value, treat people as people. This point was very touching. In an era of labeling, how much are we actually willing to explore people’s inner states rather than just quickly labeling and categorizing through external interfaces?
  • Team integration: attitude and positioning adjustment and transformation, letting go of past achievements, removing pretenses, deep communication facing contradictions and conflicts directly.
  • Life planning: coincidentally aligns with teachings from several other senior colleagues in the company - work hard, fight hard while young.
  • Intuition: an extremely important non-rule-based decision-making judgment mode. Better to frankly acknowledge the effectiveness and inexplicability of human brain neural network decision-making and use it generously.
  • Values: worldview, life philosophy, and values are components that constitute a person’s spiritual core. My original true-good-beautiful value system is completely compatible with Alibaba’s current value system, easily establishing mapping. This is actually the truly important thing, but obviously not suitable for submission as homework in technical exchange communities.

Teacher Xuannan’s lecture content has transcended the technical realm, and even programmers need to sleep. Therefore, the personal experiences and feelings from the last course can only be outlined briefly and won’t be shared here.

In summary, Baiji was indeed very powerful. I collected four answer coins throughout, got two first bloods and one second blood, exchanged for a Taobao figurine, a Taobao open source badge, plus improved knowledge level, and went home happily.

Related

2014 Year-End Summary
·1620 words·8 mins
Unable to sleep in the dead of night, spending the last hour of 2014 reviewing this year and looking ahead to next year seems quite appropriate.