Performance

Network Congestion

Network Congestion

What happens when more traffic wants to use a link than the link can carry. It is the reason the internet slows down in the evening — too many people, not enough capacity, all at the same time.

Network congestion occurs when the demand for bandwidth on a particular link exceeds its capacity. Like a motorway at rush hour, the road has a fixed number of lanes; when too many cars arrive at once, everyone slows down. On a network, the "cars" are packets, and when they arrive faster than a link can forward them, they pile up in buffers, wait their turn, and — if the queue overflows — get dropped entirely.

What congestion does to your connection

SymptomWhy congestion causes it
Lower throughputYour share of a saturated shared link shrinks as more users compete
Higher latencyPackets wait in queues before being forwarded
Packet lossWhen buffers overflow, excess packets are discarded
Buffering / lagReal-time apps suffer most from the added delay and loss

Where congestion happens

Congestion can occur at several points, and identifying which one matters for the fix:

  • Inside your home: too many devices saturating your own connection — a big download starving a video call. This is the one you can directly manage.
  • The local node / last mile: in shared technologies like cable, neighbours share capacity at the local node, so the whole street can slow at peak times.
  • Provider backhaul: if the link from your node back to the core is undersized, congestion builds during busy hours.
  • Peering and transit: a congested interconnection between networks can slow traffic to specific destinations.

How networks respond

The internet has built-in congestion control. TCP constantly probes for capacity and backs off when it detects loss or delay, so flows automatically slow to share a congested link rather than collapsing it. The downside is bufferbloat: oversized buffers absorb the overflow but add large, variable latency, which is why a saturated connection can feel laggy even while still moving data. Modern queue management — SQM and QoS — fights this by keeping queues short and prioritising latency-sensitive traffic.

Congestion vs throttling

It is easy to mistake one for the other. Congestion is accidental — a shared resource simply runs out at peak demand, and it clears when demand falls. Throttling is deliberate — a provider intentionally limits speed. The tell is timing and scope: congestion follows the clock and hits everyone on a shared link during busy hours, while throttling tends to target a specific customer, plan, or traffic type regardless of time. Comparing a peak-hour and an off-peak speed test on a wired connection is the simplest way to spot congestion.

Related Terms

More From This Section