Hi Curious Heads! 👋
If you have followed the series, you would have seen how quickly we fall into issues as we scale…
First one thing struggles to keep up, and shortly it is all a forest fire 🔥 in our VPCs!
To mitigate this, we need to have an idea of growth rates 📈 and point of failures ❌ to ensure we can scale correctly as per need and not waste our money 💸 to random scaling just to keep things running…
What do we call this step?
Ans: Back-of-the-envelope Estimations 🧾
Why?
Because they are so simple 💡 that we can do them on the go — over an envelope ✉️ or even a paper napkin 🍽️ once we’re good enough on our numbers and thinking 💯.
Let’s Practice This
Let’s take a simple, well-known app to use as a reference —
YouTube 🎥🙂
💡 What are we dealing with here?
Entities (oversimplified):
- Users
- Videos
- Comments
Types of Data:
- User Content
- Users Metadata
- Video Metadata
- Comments Metadata
- Global Distribution
- Video Analytics (likes/views/etc.)
🔁 Translate Architecture to Needs
| Component | Resource Need |
|---|---|
| Content Distribution | Network Bandwidth 🌐 |
| User + Comments + Video Metadata | Database Storage 🗃️ |
| Videos | Object Storage 📦 |
Now comes the fun part — breaking things down 💥
🧠 Memory Estimations (Storage)
User Table
- One record = 40 Bytes
- 1 Million Users →
1,000,000 x 40B = 40,000,000 Bytes ≈ 40MB
User Content
- One record = 80 Bytes
- If each user posts 10 items →
1M x 10 x 80B = 800,000,000 Bytes ≈ 800MB
And we haven’t even included indexing, caching, or relational tables yet!
📶 Bandwidth Estimations (Traffic)
Let’s assume each video view streams 2MB data on average.
If we have 500K Daily Active Users (DAU), each watching 5 videos/day:
- 500K x 5 x 2MB = 5,000,000 MB = 5 TB/day 😳
That’s 150 TB/month just in outgoing traffic — and that’s just views, not uploads!
⚙️ System Load Estimations (Throughput)
- 1M Users → ~500K DAU
- One server can handle 50K concurrent connections
So:
500K / 50K = 10 servers minimum 🖥️
Add redundancy, spikes, and background jobs and the number easily doubles.
🎯 Why It Matters?
These back-of-the-envelope estimations don’t need to be perfect — but they give you a gut-check framework to know what’s coming your way.
They help you:
- Avoid over-engineering early 🛠️
- Plan for realistic resource costs 💵
- Communicate ideas clearly to teams or stakeholders 🧑💻
Next time you think “Will our system handle this?”,
Don’t panic — pull out that napkin and do the math 💯🧾
See you in the next one!
– Anmoldeep 👨💻