Why do you want precise time?
Unless you work at a government lab, time is probably a means to an end.
People want one stock trade to be timestamped ahead of another. They want a distributed database to agree with itself about what happened first. They want a log they can hand to somebody who must know when things happened. Precise time is what those things cost — seldom the thing you set out to buy.
That sounds like a pedantic distinction and it is not, because it decides several things that follow.
It explains where the requirement comes from
Timing requirements in a datacenter usually do not originate in engineering. They arrive from a regulator, a counterparty, an auditor, or a contract — and they arrive as a number, without the reasoning that produced it.
Which means the number is often the wrong question to start from. “We need 100 microseconds” is an answer to something, and the useful move is to find out what. Whether the answer is sequencing, elapsed time, or a defensible record changes what you should build, and those three have very different costs.
It explains why “how accurate?” is the wrong first question
If precise time is a means, the first question is what it is a means to. Two systems quoting the same accuracy figure can need entirely different things:
- A system that must sequence events needs its clocks to agree with each other. It may not care much where the group sits relative to UTC — see when agreement matters more than being right.
- A system that must produce a defensible record needs traceability to a reference, which is a paperwork problem as much as an engineering one.
- A system that must measure durations needs frequency stability and possibly no timescale at all — see elapsed time.
Those are three different purchases. Asking “how accurate do we need to be?” before asking “accurate at what?” is how organizations buy the wrong one.
It explains why the work is invisible when it goes well
Nothing about a correctly timestamped network packet looks like anything. The entire return on this work is the absence of a problem: an audit that goes quietly, a reconciliation that matches, an incident timeline that makes sense.
That is worth knowing before you go asking for budget, and it is why the questions on this site are framed the way they are — as what to ask a vendor rather than as what to admire about a clock.