Smokeping
Updated

xTom Smokeping

Network latency and packet loss,
measured every second.

19 nodes in the xTom network send one ICMP echo request per second to 29 targets. Each graph shows the median round-trip time, its spread and the packet loss, from the last 3 hours to the last year.

19
Nodes
29
Targets
680
Measured paths
1.2%
Packet loss, last hour

Measurement method

How the figures on this site are collected and how to read the graphs.

What does xTom Smokeping measure?

Round-trip time (RTT) and packet loss between 19 nodes in the xTom network and 29 targets. Each node sends one ICMP echo request per second to every target assigned to it. For each minute it records the number of probes sent and lost, and the minimum, maximum and 10th to 90th percentiles of the round-trip times.

What is Smokeping?

SmokePing is an open-source latency monitoring tool written by Tobias Oetiker. It plots the distribution of round-trip times as a grey band, the "smoke", around the median. This site draws the same kind of graph with xTom's own implementation, which probes every second and stores one-minute percentiles.

How do I read a graph?

The line is the median RTT of each interval. Its colour encodes packet loss in that interval: green for 0%, light blue up to 1%, blue up to 3%, violet up to 6%, orange up to 12% and red above 12%. The light grey band spans the 10th to 90th percentile of the samples and the dark grey band the 30th to 70th. An interval with no line has no data from the node. A red mark on the top edge means every probe in that interval was lost.

How long is data retained?

One-minute samples are kept for 30 days. They are averaged into 5-minute intervals kept for 365 days and 1-hour intervals kept for 365 days. Graphs are available for the last 3 hours, 1 day, 7 days, 30 days, 90 days and 1 year.

What do the node states mean?

The state reflects a node's packet loss averaged over all of its targets during the last hour. Healthy is under 1%, Degraded is 1% to 2.5% and High loss is 2.5% or more. Offline means the node is not connected to the collector.

What is a fallback address?

Some targets have a second address. If the primary address does not answer for 30 seconds, the node probes the fallback address from the next full minute and the graph marks those intervals. It returns to the primary address after 3 minutes of replies without loss. In the latency matrix, a corner mark shows that a node is using a fallback address now.

Who operates this site?

xTom. The measuring nodes are xTom's public looking glass servers.