Web & Networking › Networking Fundamentals
Latency vs Bandwidth
How long data takes to arrive vs how much can flow at once.
Also known as: latency, bandwidth, throughput vs latency
Two different measures describe how well a network connection performs, and people often mix them up.
- Latency: how long it takes for data to get from A to B (or for a request to get a response). Measured in milliseconds.
- Bandwidth: how much data can be sent per unit of time. Measured in bits per second (Mbps, Gbps).
A useful picture: a pipe. Latency is how long a drop takes to travel through. Bandwidth is how wide the pipe is, and how many drops flow at once.
A fat pipe can still be slow to start:
satellite link: 100 Mbps, but 600 ms latency
A thin pipe can be quick to respond:
old fiber line: 10 Mbps, but 5 ms latencyWhich one matters
| Task | Limited by |
|---|---|
| Downloading a large file | bandwidth |
| Streaming video | bandwidth (and steady throughput) |
| Loading a web page with many small requests | latency, since each round trip adds up |
| An API call or database query | latency |
| Gaming, video calls | latency (and its variation, jitter) |
A page with 50 sequential requests over a 100 ms connection spends five seconds just waiting, however fast the link.
What drives latency
- Distance. Light in fiber is not instant. A trip across an ocean takes around 100 ms or more round trip.
- Number of round trips: connection setup (TCP and TLS handshakes) takes several before any data flows (TCP handshake, round-trip time).
- Congestion and queuing, routing, and processing time.
Practical consequences
- Reduce round trips: batch requests, use HTTP/2 or HTTP/3, keep connections alive (keep-alive).
- Move data closer to users with CDNs and regional deployments (CDN).
- Avoid chatty designs, such as one call per item.
- Compress and send less when bandwidth is the limit.
- Test on slow, high-latency connections, not just your office network (throttling in DevTools).