System Primer

· 1075 words · 6 minute read


  • Systems are everywhere: human bodies, the global economy, cloud computing, corporations, etc. cloud computing, corporations, etc.
  • System behaviors are often counter-intuitive. For example, destroying parts can significantly benefit the system (think about the immune system killing cancerous cells). Too many good things can hurt the system (many lottery winners end up with a worse life with poor management). It is possible to do more with less (small startups often perform much better than large companies).
  • Systematic analysis can help discover better paths than the default in real life. For example, the modern society overwhelms us with more choices. If you look at the bigger picture, however, we are connected in a way that promoting less is impossible (no one can materialize gains), even when less is much better for individuals. Thus, we should manage our own system rather than expecting society to hand out such solutions (easier option isn’t necessarily good).

Intro 🔗

Humans have tens of trillion cells (so more than 13 0s). The number of bacteria cells that live inside the body and play essential roles in health (a.k.a., microbiome) is at a similar magnitude. We as humans don’t care, or even notice, about individual cells dying, and it is estimated that we lose 50 to 70 billion cells every day (and generate as much of the new ones). For example, red blood cells have a lifespan of a mere 120 days. Overall, like the Ship of Theseus, we are still who we are (at least, we think so), even when every part of “us” is replaced.

Human System

Similarly, I want my Cloud (e.g., Gmail, Google Drive) to function seamlessly even when individual machines go wrong. Even if each machine has a 5-year lifespan with no fault in between (which sounds unrealistically optimistic), I’ll see one machine failing every hour in a cluster with 50k machines. The number will worsen as we include different failures like network, power, etc. Still, even if my server is faulty, I don’t want my Cloud to stop working until a mechanic fixes the particular machine. So, the system should handle frequent minor failures and hide them from outside view. Obviously, it has a lot to learn from the human system that handles staggeringly diverse types of failures (even DNA replications can fail) and provides resiliency (most of the time).

In sum, the whole is larger than the sum of its parts. You cannot know the whole by inspecting each part deeply. Often, the system performs better when individual parts perish, showing their interests aren’t necessarily aligned. For example, we want our immune system to terminate cancerous cells. Similarly, it can be much cheaper to terminate faulty machines than to fix an issue, even when the fault isn’t from that machine. Conversely, some actions can only benefit parts while hurting the system. For example, promoting managers by the number of reports can benefit employees who are good at expanding the team, but the overall company system will suffer from inefficiency. Overall, good behaviors for parts can be toxic for the system, and bad ones can benefit the whole.

The less code, the better! 1 point for adding a line of code, but 2 points for deleting a line. Bloatware is the devil. – Elon Musk

About 20 years ago, I found the interaction of parts frustrating as it made abstract math problems too difficult (e.g., NP problems). After two decades, I’m more intrigued than frustrated by the system as the pattern rhymes everywhere in real life. So, I’m dedicating this page (and its subpages) to touch on various topics from the first principle, and the initial focus will be on Physiology, Finance, and Distributed Systems.

Why Care? 🔗

We are bombarded by a continuous stream of temptations to try A, try B, and so on. It is always compelling as they all sound like a great solution to dramatically improve my life. My purchase is someone else’s revenue, and that’s the basis of modern society. But what if it is best not to consume? For example, a growing number of studies show the benefits of fasting. Thinking more, it is likely we, as Homo Sapiens, evolved to thrive when food wasn’t abundant and fasting was common practice. However, no one’s incentivized to promote less consumption! More food causes more illness, and more illness welcomes more medical intervention. As a result, the food industry is happy, and the pharmaceutical industry is happy. They can spend more money on ads, so the media industry is happy, too. But are we happy with the outcome?

The above example hints at the power of systematic analysis. Even when there’s no adverse player, we are often doomed to poor outcomes, and we need system analysis and proper intervention to get better results. Sometimes, it means rejecting society’s default options, like overeating or overdosing on medicine. Of course, it isn’t suggesting we should return to the prehistoric era, where individuals were powerless in front of continuous violence and diseases. Instead, we should think hard to leverage what the society offers. For example, I believe steaks, eggs, intestines, or bone broths humans consumed for thousands of years are much better options for most people (with a good combination of fasting) than sugary drinks with donuts and fat-free snacks, even when incentives to promote the ancient food are much weaker.

I directly witnessed cases where ancient lifestyles performed much better than prescriptions that doctors recommended, but health isn’t the only area to benefit from system analysis. For example, we naturally prefer expensive stocks and sell them when their prices decline (due to conformity bias), and it is no wonder why successful investors like Warren Buffett emphasize independent thinking. Likewise, we make lots of decisions in the world of systems, and if we can improve our decision-making skills, it will have compounding impacts on the quality of our lives. So, I care about the system, and probably so do many others.

You’re neither right nor wrong because other people agree with you. You’re right because your facts are right and your reasoning is right – that’s the only thing that makes you right. – Warren Buffett

Lastly, it is important not to aim for perfection. The end goal is to apply better principles in practice, and real life is inherently noisy with randomness and uncertainty. Sometimes, the outcome can be worse than the default, even when the expected value is much higher, but that’s also part of life.