How do we introduce an hour as a time unit? Historically, people measured time with the help of naturally recurring phenomena (say, a day). Then a day was split into smaller chunks (an hour, a minute, a second). Suppose we define an hour as 1/24th of a day. My questions is, how would we split a day into 24 equal parts? How can we be sure that they last the same amount?
Take the length for example. We may take anything as our 'length standard' and then directly compare its length to whatever object's size we're trying to measure. We may also split this 'unit lenght' into smaller divisions that will be more or less equal (again, because we are able to compare things with our own eyes).
With time, I can't think of any obvoius way to assess whether there is a substantial difference between two time intervals, especially if they are relatively big. So what (not overly fancy) experiment can we use to determine how long an hour (1/24th of a day) is? The same goes for minutes and seconds.
 A: Short:
Use a pendulum.
Better still, take advantage of the self timing aspects of Foucalt's pendulum.
Detail:
A "Fuocault's Pendulum" consisting of a suitably massive, suitably dense weight suspended on a suitably long, suitably thin wire will serve as a pendulum with adequately constant period of oscillation and with a duration of operation in excess of a day. Counting cycles will then allow division as desired up to at least as small as a half cycle of oscillation.
IF you have an accurate 24 hour period available the suitably compelled can count pendulum cycles over this period and then subdivide as desired.
However, as a bonus, the line of swing of the pendulum precesses with time with a rate related to the latitude of the site - no precession occurs at the equator. If latitude is known then the period of one full rotation of the line of swing is able to be calculated - so the time between any number of of swings can be calculated.
See Physics and maths explained,
General overview -  Brittanica - Foucault's pendulum and
Wikipedia - many
Values for "suitably" are 'to be established' but I have seen a steel mass of a few 10's of kg, swinging on a piano wire maybe 15 metres long with a duration of operation of many days. (In a stairwell of the Physics building At Auckland University in New Zealand - probably in 1969! :-) ).
Period of oscillation is $2\pi\sqrt{(l/g)}$ - so for a 15m wire, about 7.695 second.
To obtain a desired period, rearranging gives
$L = (t/(2\pi))^2×g$
eg For a 6 second period (1/600 of an hour) $L \approx 8. 937$ metre.
A: There's a many ways to re-define 1 hour. If you can deduce Earth rotation angle by looking at the stars/sun movement in the sky, and their respective inclination angles, you can say that  1 hour is $15 {}^\circ$ of Earth angular rotation.
Or you can measure time by distance needed to walk by foots for reaching some target place. For example, approximately 1 hour will pass while you will walk $5~km$ distance.
Yet another way is to use thermodynamic properties of matter. You can define 1 hour as time needed to melt ice cube of $m$ mass completely under direct sunlight in summer, as per
$$ 1~\text{hour} = m_{ice}\times H_{fus} \times P^{-1},$$
now solve equation for $m_{ice}$, substituting water's $H_{fus} = 333~J/g$, and solar irradiance power $P$ based on your location, time of day/year and cube irradiated area. (On winter you can employ a reverse process,- mass of liquid water $m$ needed to freeze it into ice completely in the duration of 1 hour.)
In general,- any  process fits for describing 1 hour as long as you can map process with duration and have abilities to measure progress.
A: I believe if you know the start and end of your day, you can just start walking on a plane surface try to keep the same speed $v_h$. At the end of the day count the km you walked. Say it's 120 km, then you know that 1 hour is exactly you walking 5 km with a speed $v_h$.
You can also, instead, use earth's rotation. You know well that the earth needs 24 hours to make a full rotation on it self. While it maintains the same speed, and knowing the distance between the earth and the sun $=d$, you can measure the sun's displacement needed for one hour, i.e. $$D_{\text{1hour}}=\frac{2\pi d}{24}$$
Whenever the sun moved by a distance $D_{\text{1hour}}$ then you know an hour has passed.
The earth's surface speed is $$v_\oplus=\frac{2\pi R_\oplus}{24\text{h}}$$
and so the hour is defined as
\begin{align}
\text{1 hour} &= \frac{D_\text{1hour}}{v_\oplus}\\
\end{align}
A: It is possible to define a time unit absolutely, in terms of frequency for instance.
(Of course understanding that this is the duration of a second in the inertial frame of reference - other frames of reference can experience a different duration between the start and end events of that same second.)
"The second, symbol s, is the SI unit of time. It is defined by taking the fixed numerical value of the caesium frequency ∆νCs, the unperturbed ground-state hyperfine transition frequency of the caesium 133 atom, to be 9192631770 when expressed in the unit Hz, which is equal to s−1."
https://en.wikipedia.org/wiki/SI_base_unit
For lengths, typically time is then used to define a length:
"The metre, symbol m, is the SI unit of length. It is defined by taking the fixed numerical value of the speed of light in vacuum c to be 299792458 when expressed in the unit m s−1, where the second is defined in terms of ∆νCs."
Now you may wonder, meters are defined using seconds? Doesn't this create a loophole?
Why is the meter considered a basic SI unit if its definition depends on the second?
A: Build a device an adjustable oscillator which oscillates roughly 24 times per day. Over the course of weeks or months adjust the oscillator faster or slower if you find it is ticking a little too slow or a little too fast. If you get it just right then you can leave it at that frequency. But you'll likely find that over time the hour ticker gets out of sync with days. So you can continuously adjust the hour ticker so that it ticks 24 times per day.
This is a rudimentary phase locked loop with a frequency divider.
A: Use a water clock, where time is measured by the water collected in a transparent container.
Edit: if the water flow (e.g. as drop by drop) is constant, the water height on the glass container can be divided into 24 parts with a ruler.
