Why job postings always ask for large-scale traffic
Hello, I'm Seungjae Lee. While preparing to change jobs I read a lot of job postings, and the preferred qualifications for backend roles almost never left out experience with large-scale traffic.
At first I took it as boilerplate, but as I thought about which company I wanted to join, I found myself arriving at the same criterion. Today I want to sort out why that line is always there, and why I also ended up looking for places where real traffic flows.
Development gets hard once scale starts to grow. An API that gets 100 requests a day runs no matter how you write it, but when tens of thousands of requests arrive per second, connections run out, requests trying to update the same row collide, and a few milliseconds of latency pile up into an outage.
Queries are no different. On a table with 10,000 rows you hardly notice a missing index, but at a billion rows the same query takes minutes or never finishes. Not a single character of the code changed, yet the problem to solve is a different one.
That is why I think scale is the biggest variable in how hard development gets.
Scale has two axes: lots of requests, or lots of data.
With lots of requests you start asking how big the connection pool should be, what to keep in the cache and when to evict it, and how far to take locks. With lots of data you weigh how to split tables, how much a single query reads, and whether the batch finishes on time.
Having just one of the two is enough. Either way you can no longer trust the framework's defaults, and you have to look at what happens underneath. That is when you start thinking things through to the deep end.
What if you have neither? Following the basics is enough to keep things running.
Spring takes care of the configuration, JPA writes the queries, and the database optimizer finds a better execution plan than the query I wrote. That is proof that languages and frameworks are well built, and that in itself is a good thing.
But for a developer it also means there is less to think about. N+1 queries go unnoticed when there is little data, and responses stay fast even if you forget an index. When problems never surface, there is no reason to dig in.
The organization I worked for was cost-sensitive, and while traffic was low, data kept piling up.
So I trimmed the batch down. A batch that loads more than 60 million rows every day now runs on about one dollar a month. Choosing what to cut and what to defer within the given resources was fun, and there was real engineering joy in it.
But the limits were clear. With the options fixed from the start, my thinking never got as far as "what is the right architecture" and stopped at "how do we survive inside this."
The engineering impact was small too. In an organization that spends hundreds of thousands of dollars on infrastructure, the cost you can cut through optimization also runs into tens or hundreds of thousands, but in an organization that runs on little money from the start, the ceiling on what you can save drops to match. A batch that runs on one dollar can never save more than one dollar, no matter how much you trim it.
Compared with the previous case, without scale there is no reason to think hard, and with scale but no options the range of thinking narrows. The direction differs, but the regret of lacking depth was the same.
Some will argue that services with complex domains are hard too, and I agree. In domains with tangled rules like settlement or payments, the core is how you split the model and where you draw boundaries, and that also takes skill.
But I think it is a different kind of difficulty. The difficulty of a domain comes from understanding people and rules, while the difficulty of scale comes from understanding the limits of a system. This post is about the latter, and that is the depth I want to build.
There is one more reason traffic matters: you cannot create it on your own.
You can study framework internals or index structures from books and docs, and build them in a side project. You can imitate requests with a load testing tool, but real users' traffic surges at times you didn't expect and brings data you didn't expect. You can only meet that kind of traffic at a company that has it.
That is probably why job postings prefer people with large-scale traffic experience. And so, for me, the company I want to join came to mean the same thing as a company with a lot of traffic.
And by that criterion, I have just joined an organization with more than 6.5 million MAU. I don't know yet what I will be working on, but I am dreaming of excellence in front of that traffic.
Thank you for reading.