How it works
Congestion.watch shows where the internet's long-distance networks are slow right now. It is for home users, network engineers and journalists alike, anyone asking “is it me, or is the internet slow?”
Is it me, or the internet?
The live map answers the second half of that question. It watches the long-distance networks that carry traffic between cities and continents, the ones your own internet provider connects through. It does not measure your provider, your router or your Wi-Fi.
- The map says operational, but things are slow for you? The cause is closer to home. Check your Wi-Fi, your provider, or whether the one service you are using is having trouble.
- The map shows congestion between you and what you are using, say Europe to Asia while you are on a call to Singapore? That is probably what you are feeling, and it is not something you can fix from your end.
How we measure
Probes around the world run MTR traceroutes through the major transit networks every few minutes. Each router hop is placed on the map using the location code in its hostname. The route is then split into city-to-city links, and each link is attributed to the network that operates it.
Each link's latency is compared with its own normal level, not with a fixed number, so a long ocean crossing is not mistaken for a problem.
- Slowdown means more than 15 ms over normal, or at least 1% packet loss introduced on that link.
- Congestion means more than 40 ms over normal, or at least 5% loss.
- A problem seen by only one of several probes is shown one level lower, and drawn dashed on the map, until another probe confirms it. It does not count towards a region's or network's status.
What the statuses mean
With hundreds of links measured, a few are nearly always slow somewhere. Regions and networks are therefore judged by how much of them is affected.
- Operational means nothing beyond the usual handful of slow links.
- Minor issues means at least one confirmed congested link, or 3% or more of links confirmed slow.
- Major issues means at least three confirmed congested links that make up 5% or more of the links, or 20% or more of links affected.
An issue groups the slow or congested links of one network that meet at the same city, so “Lumen congestion around Malmö” stands for every affected Lumen link out of Malmö.
Avoiding false alarms
- Routers often limit or delay replies to traceroute while forwarding traffic normally. Loss or delay counts only if it continues to every later hop.
- A link that appears faster than light in fibre has had its latency shifted by MPLS tunnelling. That latency is moved back to the link it belongs to.
- Some networks (GTT and Lumen among them) hide the routers inside their network, so a whole continent can look like one hop. Where the network allows it, those stretches are traced separately and drawn through the cities they really cross. A hop that no single cable could make (London straight to Denver) is not a real link and is left off the map, as is remote peering onto a distant internet exchange.
- Load-balanced paths and silent hops are not counted as route changes. A route change requires the sequence of networks or cities to change.
Submarine cables
Links that cross an ocean are drawn along a plausible submarine route, chaining cable systems where no single one connects the two cities. Which cable a network actually uses is not public, so no cable is named, and the route shown is an approximation. Stretches of cable with an active fault (from Submarine Cable Status) are left out until repaired.
Route changes
Each route change gets a likely cause, either congestion (the old path crossed a degraded link), outage (heavy loss on the old path), maintenance (an overnight change with no prior degradation), or unknown. It also gets a verdict, comparing latency before and after the change.
Glossary
- Latency
- How long data takes to make a round trip, in milliseconds (ms). Lower is better. Light in fibre needs about 1 ms per 100 km.
- Packet loss
- The share of data packets that never arrive. Even 1–2% makes calls choppy and downloads slow, because lost packets must be sent again.
- Congestion
- A link carrying more traffic than it can handle. Data queues up, so it arrives late or is dropped. Here, that means more than 40 ms slower than normal, or 5% or more packet loss.
- Slowdown
- A link that is noticeably slower than usual but still working, meaning more than 15 ms over its normal latency, or 1% or more packet loss.
- Normal latency
- What a link usually measures, taken as the fastest quarter of its recent measurements. Each link is compared with its own normal, not with a fixed number.
- Network (carrier)
- A company that carries internet traffic between cities and countries, such as Lumen, Arelion or NTT. Your own internet provider pays networks like these to reach the rest of the world.
- AS number
- The ID a network uses on the internet, like AS3356 for Lumen. Large companies often run several, which this site groups under one name.
- Tier 1
- The largest backbone networks. They reach the whole internet without paying any other network for transit.
- Link
- One stretch of a network between two cities, for example Lumen between London and New York.
- Issue
- Slow or congested links on one network that meet at the same city, grouped so you can see where a problem is.
- Region and network status
- Operational means nothing beyond the usual handful of slow links. Minor issues means some confirmed congestion, or 3% or more of links confirmed slow. Major issues means several confirmed congested links making up 5% or more of the region, or 20% or more of links affected. Unconfirmed readings do not count.
- Route change
- When traffic between two places starts taking a different path, through other networks or other cities. Networks do this to avoid trouble or during maintenance.
- Probe
- A server that runs the measurements. Probes in many countries trace routes across the networks every few minutes.
- Unconfirmed
- Only one probe has seen this problem so far. It is drawn dashed on the map until a second probe confirms it, and shown one level lower when other probes cross the same link without seeing it.
- Traceroute (MTR)
- A test that lists every router a packet passes through and times each one. This site runs MTR, a traceroute that repeats to measure loss as well.
- Alert
- A record of something that went wrong at one moment, such as a congested link, a slow or lossy path, an unreachable destination, a route change. Alerts stay in the list after the problem clears; active issues do not.
- Path
- The whole route from one probe to one destination, through every network and city on the way. A link is one stretch of a path.
- How latency is corrected
- Routers often answer a traceroute late, or not at all, while forwarding traffic normally, and some networks hide the routers inside them. Delay or loss counts only when it carries on to every later hop, and latency that appears on the wrong stretch is moved back to the one it belongs to.
- Submarine cable
- Fibre on the sea floor that carries traffic between continents. A cable fault forces traffic onto longer routes.