lunch with the GOAT @striver_79 today🥰

Developed a URL shortener. Stack: Go + Fiber + Redis + Docker. Added Security Features:.- Rate limiting.- URL validation .- Domain loop prevention.- Redis TTL

Join millions who have switched to Grok.

Deploying new features also became less risky. Kubernetes' rolling updates allowed Netflix to gradually roll out new versions of their services, monitoring for issues and automatically rolling back if something went wrong , minimizing downtime for their massive user base.

And when the initial rush subsided? Kubernetes could automatically scale back down, optimizing resource usage and saving costs. No more over-provisioning servers "just in case.".

Imagine the "Stranger Things" season premiere. Millions of people log in at once. With Kubernetes, Netflix could automatically spin up hundreds or even thousands of additional instances of their relevant microservices to handle the load, without any human intervention.

This is where Kubernetes (though Netflix actually pioneered some of the underlying concepts before Kubernetes existed) provided the answer. By containerizing their microservices and orchestrating them with a system like Kubernetes, they could automate scaling.

Netflix, with its massive and unpredictable traffic spikes, needed a better way. They weren't just deploying one monolithic app, they had hundreds of microservices responsible for everything from recommendations to video encoding.

Think back to the early days of streaming. A new season of a popular show would drop, and BAM! The servers would get hammered. Manually adding and managing servers to keep up was a constant, stressful fire drill.

Ever had a website slow down during a flash sale? . Or maybe an app that just couldn't handle the sudden surge in users? . Companies like Netflix faced this exact scaling nightmare before fully embracing Kubernetes.

For celebrities with millions of followers, a push model would break the system. So, for them, they use the pull model. Their fans' feeds pull the posts on demand.

Most big platforms use a hybrid model. For regular users, they use the push model to get the post to a few hundred or thousand followers quickly.

Push Model (Fan-out on Write): When you post, the system immediately pushes your post into the feeds of all your followers. This is lightning-fast for the user, but a nightmare for the system if you have a million followers.

Pull Model (Fan-out on Load): When you open your feed, the system pulls the latest posts from everyone you follow. This is simple, but it can be really slow if you follow thousands of people.

The core challenge is simple: how do you get a post from one person to all of their followers? There are two main ways:.Pull Model (Fan-out on Load).Push Model (Fan-out on Write).

Every time I scroll Twitter or Instagram, my mind is blown by the engineering. Here is what it takes to design a system to handle billions of posts a day:

Little glitch on the @ycombinator startup directory: filters break on first load, only work after refresh. Need a good frontend engineer to fix this? @garrytan 👀

This is what goes behind the notifications we receive everyday 🤯

These next 4–5 months aren’t your final push, they’re one of the most impactful pushes you’ll ever make for your career. Stay locked in.

So, the final architecture looks like this: a client sends a request to a load balancer, which forwards it to an API server. The server either generates a new short URL or looks up an existing one in the database and sends a 301/302 redirect

The fix for collisions is a bit messy. You might have to add a random string to the long URL before hashing, or just retry with a different hash. This adds complexity, which is why the base62 counter approach is often preferred for its simplicity and reliability