NTP and NTS
Is this service paid, and do I need to register?
No. time.nerzen.com is public and free; no registration, key or permission is required. Entering the server name in your client is all it takes.
How many NTP servers should I configure?
At least four. With one server you cannot detect an error; with two, you cannot tell which one is right when they disagree; three drop to two when one is unreachable. Four sources are usually enough to reach a decision even if one goes bad. Use time.nerzen.com as one of those sources.
What does stratum mean? Is stratum 2 bad?
Stratum is not a quality grade; it is a measure of distance from the primary clock. A server connected directly to an atomic clock or a satellite receiver is stratum 1; one that gets its time from that is stratum 2. Our server is a stratum 2 connected directly to twelve stratum 1 sources. A well-connected stratum 2 can give you more accurate time than a stratum 1 you reach over a long, poor path; what decides it is the network path.
What is the difference between NTP and SNTP?
They use the same packet format. A full NTP client (chrony, ntpd) measures several sources continuously and slews the clock smoothly by correcting its frequency. An SNTP client (most embedded devices, timesyncd) asks a single source and sets the clock to it. Our server answers both in the same way.
Do you support NTS?
Yes. The key exchange takes place over 4460/tcp with TLS 1.3, and time responses over 123/udp. See the NTS section for setup. We also authenticate half of our own upstream sources with NTS.
Do you support IPv6?
Yes; time.nerzen.com has both A and AAAA records, and NTP and NTS work over both protocols.
How do you handle leap seconds?
The standard way: the leap second announcement is passed to clients in NTP's own field, and the second is inserted by the kernel at the end of the day. We do not apply a "smear" that spreads the second over several hours; this means the server can be mixed with other non-smearing sources without problems. There is an international decision to discontinue leap seconds by 2035 at the latest; until then a new leap second can still be inserted.
Does Türkiye observe daylight saving time? Does it affect NTP?
Türkiye has used UTC+03:00 all year round since 2016. NTP only carries UTC anyway; time zone and daylight saving conversion are entirely your device's job.
How reliable is the offset measurement on this page?
It is a rough indicator. The browser asks for the server's time six times over HTTPS and uses the measurement with the shortest round trip; the error bound is half of that round-trip time. It reliably shows a drift of tens of milliseconds, but not single-digit milliseconds.
If I leave the page open, will the displayed time drift?
No. The time is displayed by adding the offset obtained from the server to the browser's monotonic counter, not to your device's wall clock, and the measurement is refreshed every five minutes.
Can I use it in production?
Yes; Nerzen's own network backbone takes its time from this server too. Even so, no production system should depend on a single time source; configure it together with several independent sources.
Does my server or VPS on the Nerzen network need any extra configuration?
No, use the same name. Because you are in the same data centre as the server, the round trip is under one millisecond and you get the highest accuracy.
Sources and accuracy
Where do you get the time from?
From twelve stratum 1 servers in six countries: national time laboratories and satellite (GNSS) disciplined servers. Which one is in use at the moment, and the measured offset of each, is shown live in the table on the home page.
Why so many sources?
For reliability. The sources are spread across different organisations, countries and network paths; if one organisation gets it wrong, the line to one country goes down, or one server comes under attack, the rest still form a majority and the clock is unaffected.
How close is the server's clock to UTC?
The upper bound we publish is "half the root delay + root dispersion", shown live on the home page; under normal conditions it is a few milliseconds. The actual offset is usually far below that, but this is the bound we can prove.
Do you have your own atomic clock or GPS receiver?
No. This is a stratum 2 service and we say so plainly. We take time from stratum 1 servers that have their own primary clocks, cross-check it and distribute it.
What happens if all sources are lost at once?
The server carries on on its own oscillator; because the measured frequency error is corrected, under normal conditions it drifts by less than one millisecond per hour. Meanwhile the published root dispersion grows, which tells your client "my uncertainty is increasing". When the sources return, synchronisation resumes on its own.
Usage rules
- Query frequency: use your client's default interval (64–1024 seconds). Queries from the same address more often than once per second hit the rate limit.
iburstat startup is fine; continuousburstand settingminpollbelow 4 are not.- Embedding in a product: let us know before you set it as the default time server in a device or piece of software that you sell or distribute.
- Large numbers of devices: if you synchronise hundreds of devices from behind a single NAT address, set up your own server inside and point only that one at us.
- Warranty: the service is provided on a best-effort basis, with no service level commitment. Outage and maintenance information is published on the status page.