Node HandbookNODE HANDBOOK中文
Nodes · performance

Why a node is fast at noon and unusable at night

Evening slowness is usually a capacity question at one of three points: your own connection, the route between networks, or the provider's shared capacity. Each has a different test.

Node Handbook editorial team ·

Three places congestion can be

PointHow to test itWhat the result means
Your local linkSame hour, proxy off, same targetIf direct also slows, the local connection is part of the problem
The route between networksSame hour, proxy on, a different node in another regionIf only one region collapses, the route or that region's peering is implicated
The provider's capacitySame hour, several nodes in the same regionIf all of them degrade together, shared capacity is the likely constraint

Measure the right thing

Peak-hour faults rarely show up in minimum latency. They show up as:

  • Lower sustained throughput for a fixed file.
  • Higher variance: the same transfer takes wildly different times.
  • Stalls in streaming or video calls while ordinary pages still load.
  • Retransmission-like behaviour: slow but eventually succeeding.

A single small download measures none of these. Use a sustained transfer of a meaningful size, at the hour you actually use the connection, and record both the result and the time.

A comparison that produces evidence

  1. Pick a fixed target and a fixed file.
  2. Test at 03:00 and at 21:00, same node, same network.
  3. Repeat with the proxy off at both hours.
  4. Repeat at 21:00 with a second node in the same region.
  5. Record throughput and variance, not just “fast” or “slow”.

That gives four numbers and a comparison. “It is slow at night” gives none, and cannot be acted on by you or by a provider.

What usually is not the cause

  • DNS. If names resolve, DNS is not making transfers slow.
  • Your client's rule set. Unless a rule change coincides exactly with the hour, routing is not time-dependent.
  • The node being “dead”. A node that works at noon and degrades at night is congested or throttled, not dead.

What you can actually do

FindingReasonable response
Direct traffic also degradesLocal link or ISP peak; a proxy will not fix it
One region degrades, others do notUse another region at peak, or ask the provider about routing
All nodes in a region degrade togetherShared capacity; ask the provider, or move region for peak hours
One node degrades, its neighbours do notThat server; switch within the region

Recording it for a provider

  • Time and time zone, on several days.
  • Node label, protocol, observed exit.
  • Fixed target and file size.
  • Throughput and variance at peak and off-peak.
  • The same measurements with the proxy off.

That is a report a provider can act on. “Slow at night” is not.

Local causes that look like node congestion

Local causeHow it presentsHow to confirm
Household bandwidth saturationEverything slows at the same hour, proxy or notDirect control at the same time; check other devices
Wi-Fi contention or weak signalVariance much higher on wireless than wiredSame test over Ethernet
Router session or NAT limitsMany connections slow down together; single transfers are fineReduce concurrent connections and retest
Device power savingSlowness on battery, better when plugged inSame test on mains power
Background transfersIntermittent, correlated with sync or update windowsPause them and retest

All five produce “the proxy is slow tonight”. Only the control test separates them from the node.

When the provider is the constraint

Two patterns point at shared capacity rather than at your line. First, every node in one region degrades in the same window while another region stays usable. Second, the degradation begins at a consistent local hour across several days, independent of your own traffic. Neither is proof, but together they are a report a provider can investigate.

What a provider can change is capacity and routing. What they cannot change is your last mile, your Wi-Fi, or the destination's own peak. Sending them a measurement that mixes all three produces a template reply, not an investigation.

Scope

This page describes a measurement method. It contains no provider test results and makes no claim about any specific network's peak behaviour — which varies by region, carrier and month.