Network peering and latency: what actually affects your connection
Kort antwoord
Latency isn't decided by how big a network's total capacity is, it's decided by the path your traffic takes and how congested that path is. A provider with more Tbit/s of capacity handles more overall traffic without slowing down, but that's a separate thing from how quickly a single packet gets from A to B. Peering type (public at an IXP versus a private PNI) shapes the available paths, and congestion, not distance, is the usual explanation when ping times spike at specific times of day.
Public peering (IXP) vs. a PNI: two different ways to connect
Networks exchange traffic with each other in two main ways:
- Public peering at an IXP (Internet Exchange Point), such as AMS-IX. A shared fabric where many networks connect at once, so one connection to the exchange gives broad reach to everyone else present there. It's efficient and cost-effective, but it's a shared resource.
- A PNI (Private Network Interconnect), a dedicated, bilateral link between two specific networks. Because it's not shared with anyone else, a PNI can reduce contention and provide predictable capacity for high-volume or latency-sensitive traffic between those two networks specifically.
Neither is universally "better", they solve different problems. Public peering gives broad, efficient reach; a PNI gives a dedicated, predictable path to one particular partner. A network that uses both, public peering for broad reach and PNIs where a specific route justifies it, generally has more options for keeping paths short and predictable.
Why a bigger network (more Tbit/s) doesn't automatically mean lower latency
Total network capacity, measured in Tbit/s, is a headline number, but it answers a different question than latency does. Capacity is about how much traffic a network can carry without congestion. Latency is about how long it takes a single packet to make its journey, which depends on the route it takes, the number of hops, and whether any point along that route is congested. A provider can have enormous total capacity and still hand your traffic a slow, roundabout path if the routing and peering aren't set up well. Conversely, a well-peered network with shorter, more direct routes to the networks your users are actually on will typically deliver lower latency, independent of how big its headline capacity figure is.
In short: capacity tells you how much traffic fits through the pipe, latency tells you how quickly a single packet gets through it. They're related but not interchangeable, and a bigger pipe on its own doesn't shorten the trip.
Why pings spike at specific times of day
A ping that's fine during the day and noticeably worse in the evening is usually not a distance problem, the physical route hasn't changed. It's typically peak-hour congestion: more traffic competing for the same shared capacity at busy times, often combined with bufferbloat, where queued-up packets sit waiting on an oversubscribed link instead of being dropped or prioritised. That pattern points at congestion somewhere on the path rather than at the underlying network architecture. Quality of service (QoS) or smart queue management (SQM) settings can help absorb some of that on a local network, and connecting through an ISP with better peering can reduce how much of the path is exposed to shared, congested links in the first place.
Testing it yourself
If you want to see what's actually happening on your route rather than guessing, standard network diagnostic tools (ping, traceroute/tracert, and mtr for latency and path behaviour, iperf for throughput specifically) will show you the real picture. We cover how to run and read these in detail in Traceroutes & MTRs, worth reading alongside this article if you're trying to diagnose a specific latency or routing issue rather than understand the concepts behind it.