Networking

Bandwidth, Latency & Jitter

"Bandwidth" gets used loosely to mean "how fast is my internet" — but it's really only one of several independent properties that together decide whether a connection actually feels fast. A link can have enormous bandwidth and still feel sluggish, for reasons bandwidth alone can't explain.

Bandwidth vs. throughput

Bandwidth is the theoretical maximum a link could ever carry. Throughput is what actually gets through in practice — always less than or equal to bandwidth, reduced by protocol overhead, congestion, retransmissions, and whatever else is sharing the link at the same moment. A "1 Gbps" link that measures 600 Mbps of real throughput isn't broken — it's simply never operating under the one perfect, uncontended condition its rated number describes.

Latency — a completely independent property

Latency is how long one single bit takes to travel from sender to receiver — and it has almost nothing to do with bandwidth. A satellite internet link can offer very high bandwidth while still feeling laggy for anything interactive, because the physical round trip to a satellite and back takes real time no amount of extra capacity can shorten.

Jitter

Jitter is variation in latency over time — packets arriving at inconsistent intervals instead of a steady, predictable rhythm. A connection with mediocre but consistent latency can feel smoother for real-time audio/video than one with a lower average latency that occasionally spikes — the spikes are what a listener actually perceives as a stutter, not the average number on a speed test. This is exactly why real-time media reaches for UDP in the first place: waiting for a retransmission introduces exactly the kind of irregular delay jitter describes, which is worse than simply dropping the occasional sample.

Bits vs. bytes

Network speeds are conventionally quoted in bits per second (Mbps); file sizes and download progress bars are conventionally shown in bytes (MB) — and there are 8 bits in a byte. A "100 Mbps" connection has a real-world ceiling around 12.5 MB/s, not 100 MB/s — not a fault in the connection, just two different units that look deceptively similar written down. This single unit mismatch is behind an enormous fraction of "my internet isn't as fast as advertised" complaints.

Real-world applications

• Video streaming — primarily a bandwidth problem — enough sustained throughput to keep a buffer filled matters far more than latency for a service like this.
• Competitive online gaming — primarily a latency problem — a player's connection can have modest bandwidth and still feel great, or huge bandwidth and still feel bad, entirely depending on round-trip time to the game server.
• VoIP and video calls — primarily a jitter problem — audio quality degrades from inconsistent delay long before raw bandwidth becomes the bottleneck.
• Embedded/IoT telemetry links — often bandwidth-constrained by design (cellular data costs, LPWAN radio limits), which is exactly why compact serialization formats and protocols like CoAP matter more here than they would over a home broadband connection.

Gotchas