Network Throughput Constraints
What you'll learn: How a single machine's network interface card (NIC) limits your system's bandwidth, and why adding more machines helps break through that ceiling.
The Network Card Bottleneck
Every server has a physical network interface card that connects it to the internet. This card has a maximum bandwidth—commonly 1 Gbps, 10 Gbps, or 25 Gbps in modern data centers. This is a hard physical limit on how much data can flow in and out of that machine per second.
Think of it like a water pipe: no matter how powerful your pump (CPU) or how large your tank (memory), the pipe diameter determines maximum flow rate. A single 10 Gbps NIC cannot push more than ~1.25 GB/s of data, period.
Why This Matters for Scaling
When you vertically scale (add more RAM, CPUs to one machine), you don't increase network capacity. If your application serves video streams, large file downloads, or handles massive API traffic, you'll hit the network card's limit long before maxing out CPU or memory.
Example: A video streaming service might need to serve 100 Gbps during peak hours. One server with a 10 Gbps NIC is physically impossible. You must use horizontal scaling.
The Horizontal Scaling Solution
When you add more machines, you aggregate network capacity. Ten servers with 10 Gbps NICs give you 100 Gbps total throughput. Each machine handles its portion of traffic independently, and together they break through the single-machine constraint.
This is why content delivery networks (CDNs) and large-scale systems distribute across many machines—they need combined bandwidth that no single machine can provide.
Key Takeaway: A single server's network card caps total bandwidth; horizontal scaling multiplies network capacity by adding more independent network connections across multiple machines.