@rushil_b_patel Itโs not just a problem of javascript , it used to be a problem of every programming language and this has caused many catastrophic failures in real life - more content is on this video -
https://t.co/B23V6CSZUR
Created a small video on how computers see floating point numbers .
Give it a watch and let me know the feedback or suggestions
https://t.co/B23V6CSZUR
Created a small video on how computers see floating point numbers .
Give it a watch and let me know the feedback or suggestions
https://t.co/B23V6CSZUR
2022 โ writing code
2023 โ copying code from ChatGPT
2024 โ prompting AI to write code
2025 โ reviewing AI's code
2026 โ watching AI write code
2027 โ AI watching you
we call this "career progression"
Microservices don't solve complexity.
They move it , from code to infra.
Now you need:
Service discovery โ Distributed tracing โ API gateways โ Eventual consistency
You traded 1 problem for 5.
Start simple. Break it only when you feel the pain.
What are you running?
@iikypjjyl Fair catch!
No study , it's a hook, not a stat.
The real stat that should matter is how many engineers have actually debugged a cascading failure in prod vs how many have just read about circuit breakers?
My guess? Still a pretty wide gap.
99% of engineers design systems that work.
1% design systems that survive. The difference isn't intelligence.
It's knowing these 5 failure points that nobody teaches you in interviews, but every senior engineer learns the hard way.
๐งต Thread:
Failure point #4: Hot partition
Shard DB by user_id.
One viral user generates 10,000x normal traffic.
That shard melts. Rest sit idle.
Fix: Consistent hashing + virtual nodes.Distribute by request pattern , not just by key. The worst bottlenecks are invisible until prod.
Failure point #3: Retry storms
Service dies for 10 seconds.
Every client retries instantly.
Service comes back and dies again.
Fix: Exponential backoff + jitter. Not 1s, 2s, 3s. Try 1s, 4s, 16s ... randomized. AWS uses this. Google uses this.
So should you.
Failure point #2: Thundering herd
Cache expires. 10,000 users hit your DB at the same millisecond. DB collapses.
Fix: TTL jitter. Add randomness to expiry times. Only 1 request rebuilds cache , rest wait.
One line of jitter. Entire system saved.
Failure point #1
Cascading failures
You design Service A โ Service B โ Service C.
C slows down. B queues up requests.
A's thread pool fills. Your entire system dies
Fix: Circuit breakers. If C fails N times, stop calling it.
Return a degraded response instead of dying