Latency vs. Bandwidth
Two metrics that define network performance.
Detailed Explanation
Latency is the 'delay' (the time it takes for one bit to travel from A to B, usually in milliseconds). Bandwidth is the 'capacity' (how much data can be sent per second, e.g., 100 Mbps). For web browsing, low latency is often more important than high bandwidth. A fast 10 Mbps fiber connection feels faster than a 100 Mbps satellite connection because the fiber has lower latency.
Quick Summary
Latency is how long a single round trip takes; bandwidth is how much data fits per second. For interactive web apps, latency almost always matters more, and you can't buy your way out of the speed of light.
Key Takeaways
- Latency floor is set by physics: ~30ms within a continent, ~150ms across an ocean, ~600ms via satellite.
- Bandwidth keeps growing; latency improves slowly because it's bounded by distance and protocol overhead.
- A page that loads 100 resources at high latency feels slow even on a fast pipe, round trips dominate.
- Throughput = window size / RTT. Higher latency = lower achievable throughput on a single TCP connection.
- CDNs, HTTP/3, and HTTP/2 multiplexing all attack latency, not bandwidth.
When to use it
- Choosing a CDN to put static assets closer to users.
- Bundling and reducing the number of network round trips on critical paths.
- Deciding where to host: region choice often matters more than instance size.
- Diagnosing "the network is slow", distinguish latency complaints from throughput complaints.
Common Mistakes
- Conflating Mbps with "fast", high bandwidth + high latency is still a frustrating experience.
- Optimizing image size when the real cost is 12 sequential round trips.
- Ignoring last-mile latency from mobile networks; servers respond fast but users feel slow regardless.
- Measuring on dev machines (localhost ~0ms) and being surprised by real-world numbers.
Latency vs. Bandwidth, Frequently Asked
If I double bandwidth, will my site load faster?
Maybe slightly. Most page loads are latency-bound, too many sequential requests, not too few bits per second. Doubling bandwidth helps if you're shipping huge assets; it doesn't help if your bottleneck is the chain of dependencies.
How do I reduce latency?
Move closer (CDN, edge compute), do fewer round trips (HTTP/2 multiplexing, HTTP/3, combined APIs), and overlap work (preconnect, prefetch). The biggest wins come from eliminating round trips, not making each one faster.