Network Latency
Latency is why things feel slow even on a fast connection. These guides break a round trip into its parts and give reference numbers for what delay to expect on different paths. They expose the hidden killer called bufferbloat and walk through a troubleshooting path for high ping times.
Guides in this topic
- Anatomy of a Round Trip: Where the Milliseconds GoDNS, handshakes, propagation, queuing, and server think-time — one click decomposed with real numbers.
- The Four Delay Types: Propagation, Serialization, Queuing, ProcessingThe physics and math of delay, with worked examples.
- What Latency Should You Expect? Reference Baselines by Path TypeSame-switch to satellite, VPN overhead, and cloud-region tables — the bookmarkable reference.
- Measuring Latency HonestlyICMP caveats, TCP connect timing, and application-level measures — a layered latency test kit.
- Bufferbloat: The Hidden Latency KillerWhat it is, how to test for it, and the queue-management fixes that transform a link.
- Latency vs. Bandwidth: Why the Fat Pipe Still Feels SlowThe explainer to send to anyone who wants to fix slowness by buying more megabits.
- Troubleshooting High Round-Trip Times: A Decision TreeClient, LAN, WAN, and far-end causes — with the test that isolates each branch.
Quick answers about latency
How do I lower latency, and how do I fix high latency?
First separate the two kinds: distance latency, which only a shorter path fixes, and queueing latency, which comes from a busy device or a full link and is fixable. Loaded versus unloaded latency tells them apart: if latency rises when the link is busy, the cure is the queue, not the distance.
Is high latency good or bad, and what is a good latency?
Lower is better, and consistent is better than merely low. Under 20 ms on a LAN is normal, 20 to 80 ms across a country is normal, 150 ms is where voice calls start to feel wrong, and a sudden increase from your own baseline matters more than any absolute number.
Latency vs speed vs bandwidth: what is the difference?
Bandwidth is how much a link can carry per second; latency is how long one packet takes to cross it; "speed" in everyday use means either, which is why a fast link can still feel slow. The guide on latency versus bandwidth explains why a database over a high-latency link crawls on a gigabit circuit.
How do I calculate latency, and what do p50, p90 and p99 latency mean?
Measure the round-trip time of many packets and summarise: the median (p50) is what a typical packet sees, p90 and p99 are what the unlucky ones see, and the gap between them is jitter. Averages hide the tail; the tail is what users complain about.
Why does DNS resolution have high latency, and where does web page latency go?
A slow resolver, a resolver reached across a WAN, or a cache that keeps missing. Web page latency is DNS, then the TCP and TLS handshakes, then the server's own response time, then the transfer, and a monitor that measures each step shows which one grew.
How do I improve network latency between sites and data centers?
Measure the path with a packet train between probes, remove the queueing (QoS for the real-time classes, more capacity for the saturated hop), and shorten the path where the routing allows. Communication latency between data centers is mostly distance; everything else is yours to fix.
Increased latency and latency issues: what affects latency, and how to get lower latency
Distance, queueing, the number of hops and the processing at each one affect it; only queueing is quick to fix. To get lower latency, find the hop whose latency rises when the link is busy and relieve it: QoS for the time-sensitive class, more capacity, or a shorter path.
Is 20 ms latency good? Is higher latency better or worse?
Lower is always better. 20 ms is good for anything short of competitive gaming; 9 to 13 ms on a LAN is normal; what matters more than the number is whether it is steady and whether it is your own baseline.
How do I decrease and improve network latency, and can I simulate latency to test?
Decrease it by removing queues on the path and by keeping the time-sensitive traffic in a priority class. To test what an application will do at a given latency, a WAN emulator or a Linux traffic-control rule adds delay on a test link, which is cheaper than finding out in production.
What is a latency issue, and what does it mean when DNS resolution has high latency?
A latency issue is a delay that has grown past what the application tolerates: a round trip that used to take 20 ms and now takes 200, or a path whose delay swings so much that voice and video stutter. DNS resolution has high latency for its own reasons: a resolver that is far away, forwarding through several hops, a cache that keeps missing, or a primary resolver that times out before the secondary answers. Time the lookup separately from the connection, because a slow name lookup looks exactly like a slow server to the user.
Put this into practice with SysPeep Monitor
SysPeep is network and infrastructure monitoring software for Windows. It discovers the devices on your network, polls their health and performance over SNMP and WMI, and alerts you the moment something goes wrong — all from one web console.
Download Free TrialLearn About SysPeep