---
title: "Discussion on \"Diagrams never crash, so I built a game where your system design does\" | Hashnode"
description: "Discussion on \"Diagrams never crash, so I built a game where your system design does\". I kept reading about load balancers, caches and sharding, then forgetting all of it a week later. A diagram never fails, so you never find out why any of it matters.\nSo I built Bugpit, a game where yo"
url: "https://hashnode.com/posts/diagrams-never-crash-so-i-built-a-game-where-your-system-design-does/6aff4cdcf9cae7e671c5b4d7"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Diagrams never crash, so I built a game where your system design does](https://bugpit.hashnode.dev/diagrams-never-crash-so-i-built-a-game-where-your-system-design-does?utm_source=hashnode&utm_medium=feed)

I kept reading about load balancers, caches and sharding, then forgetting all of it a week later. A diagram never fails, so you never find out why any of it matters. So I built Bugpit, a game where yo

1d ago · 3 min read · bugpit.hashnode.dev  
Read the article

-   [#system-design](https://hashnode.com/tag/system-design)
-   [#webdev](https://hashnode.com/tag/webdev)
-   [#gamedev](https://hashnode.com/tag/gamedev)
-   [#side-project](https://hashnode.com/tag/side-project)

## Written by

[Umar Shaikh](https://hashnode.com/@umars)

@umars · Mumbai, India

  
[View profile](https://hashnode.com/@umars)

## Responses 3

Markdown supported · ⌘↵ to post

[Ehsan](https://hashnode.com/@ibuildinpublic)

[1d ago](https://hashnode.com/posts/diagrams-never-crash-so-i-built-a-game-where-your-system-design-does/6aff4cdcf9cae7e671c5b4d7/comment/6b000e6a183297a9757278c2)

The write wall example with 8,000 writes a second and a 4,800 cap finally made sharding click for me. Where do people get stuck most on level 9, splitting shards or realizing the replica still eats every write?

[Umar Shaikh](https://hashnode.com/@umars)

[1d ago](https://hashnode.com/posts/diagrams-never-crash-so-i-built-a-game-where-your-system-design-does/6aff4cdcf9cae7e671c5b4d7/comment/6b000e6a183297a9757278c2#reply-6b001bd1d6ec62877fd85a30)

Glad it clicked! I don’t have enough data yet to say for sure, but from how the level is built, the replica is the trap. Most people try a read replica or a bigger database first, because that fixed the earlier levels. But every replica still has to apply every write, so 8,000 writes a second hits each copy just the same. Once that lands, the split itself is easy: 4 shards at about 2,000 writes each fits under the 4,800 cap with room to spare.

The part people don’t expect is the bill. Sharding turns out to be the most expensive line in the design, which is exactly why teams put it off as long as they can.

[DemoVexa](https://hashnode.com/@demovexa)

[7h ago](https://hashnode.com/posts/diagrams-never-crash-so-i-built-a-game-where-your-system-design-does/6aff4cdcf9cae7e671c5b4d7/comment/6b01079a33f51a2ce2f1af81)

Level 9 is a great hook because the failure is something you can watch: 22.6% of requests failing, then none once the database is split into four shards. For a launch clip, I'd record exactly that run, the read replica attempt first and then the shards, with the failure count and monthly bill on screen together. Ehsan's question shows the replica trap is what sticks, and watching the bill climb teaches the second lesson without narration.
