Ping — Round-Trip Time and Packet Loss to a Host
Ping sends small ICMP echo requests and times the replies, reporting round-trip time and how many probes went unanswered. It is the oldest and simplest network diagnostic, and it remains the first thing worth running because it separates two questions cleanly: is the host reachable at all, and if so, how long does the round trip take.
Why run it
The distribution matters more than the average, and this is where most readings are misinterpreted. Consistent 40ms is a healthy link. An average of 40ms made up of mostly 20ms with periodic 300ms spikes is a link with a problem, and the two report similarly if you only glance at the mean. Packet loss is read wrongly just as often: a small percentage sounds negligible and is not, because TCP responds to loss by backing off, so 2% loss can halve throughput. Note too that a host not responding may simply be filtering ICMP rather than being down.
Why automate it
Latency problems are frequently periodic, and a periodic problem is invisible to a manual check by definition — you run it, it looks fine, and the spike happens at 4am or on the hour when a backup runs. Continuous sampling builds the distribution rather than one sample from it, which is what makes the pattern visible: a daily spike, a weekly degradation, a slow upward drift as a link fills. It also gives you the baseline to compare against, and without a baseline "the network is slow" is an opinion.
What it gives you
The reference measurement for everything else. Sampled continuously it turns latency from an argument into a chart, and it is the cheapest way to know whether today is unusual.