You may not have asked yourself, but when it comes to installing a new HDD or SSD on your PC, there are a series of steps to follow if we want to partition them to limit the performance or capacity of these components. In SSDs this is much more controlled by its nature, but in HDDs it is very important to understand how operating systems manage to partition, since we can lose performance if we do not do it well.
We are not going to explain how to make a partition in itself neither in Windows nor in Linux since we already have articles related to them directly, but failing that we are going to deal with another topic that is equally interesting: “how do they do it”.
Do partitions on HDDs and SSDs affect performance? Are they physically attached?
They are two great questions because depending on whether we talk about HDD or SSD there is one answer or another. We will start with the latter since it is much simpler to answer and we will end with the HDDs in more depth.
In an SSD, a partition does not influence performance due to the nature of the component itself and its NAND Flash. When making a partition on an SSD the only thing we do is tell the MBR / GPT what the partition table is like and in which sectors and blocks each one begins and ends.
Obviously, the blocks are assigned to save a predefined space according to our needs and because the controller takes the same time to access some blocks or others, we will not have any loss of performance.
On the union, normally an SSD reserves whole blocks when it comes to partitioning, so it is tried not to share these blocks between two partitions, so the union as such is physical in the worst case, but does not affect anything , it is irrelevant to the controller that data is shared from two different partitions in the same block.
Why is this not true on an HDD?
Well, as in the SSD it does not affect by its nature and technology, in an HDD it is the opposite for the same reasons. As we all know, the outer parts of the cymbals rotate at higher rotational speeds precisely because of their angular velocity. This influences a term known as rotational latency , which is obviously not the same at all points of the plate.
The closer we get to its edge and the more RPM the motor and its axis have, the less latency we will have when reading the data. This directly influences the search time needed to position the spindle on the specific track of the chainring, something that helps the NCQ to calculate shorter paths.
Therefore, an operating system should be located on the periphery of a plate, something that Windows and Linux automatically achieve as long as there is no data on it before installation. When a partition occurs, the printhead marks the beginning and end on the platter, and from there continues to write, so there is a physical bond between partitions due to magnetism and shared material.
The partition table registers it and from there the HDD continues writing the new data. But, what happens if our hard disk has several plates? The tactic is somewhat different because although we have several heads, the allocation and partition can vary depending on their size or sizes.
The ideal according to the addressing mechanisms is that if the partition is feasible on a plate, there are only two heads intended for it, leaving the rest without work, but if this does not happen, the partition will be made on different plates symmetrically , for what information will be read and written in all of them at the same time dividing the capacity among them.