Stan Cao

A backend developer.

Disk Storage

1 minute read

Disk Storage

An HDD is logically divided in a number of different regions. Seeking is usually the most expensive operation on a hard drive.

  • Platter. A non-magnetic, circular storage surface, coated with a ferromagnetic film to record information. Normally both the top and the bottom of the platter are used to record information.
  • Track. A single circular “slice” of information on a platter’s surface.
  • Sector. A uniform subsection of a track.
  • Cylinder. A set of vertically overlapping tracks.

Cluster is a contiguous group of sectors, a logical concept for OS to read data.

Rather than using sectors, some OSs allowed users to store data in variable-sized “blocks.”

See the graph below for an HDD.

disk

Access Cost

The cost of a disk access includes

  1. Seek. The time to move the HDD’s heads to the proper track. On average, the

    head moves a distance equal to 1􏰇/3 of the total number of cylinders on the disk.

  2. Rotation. The time to spin the track to the location where the data starts. On

    average, a track spins 1/􏰇2 a revolution.

  3. Transfer. The time needed to read the data from the disk, equal to the number of bytes read divided by the number of bytes on a track times the time needed to rotate the disk once.

Reference

[1] Disk-Based Algorithms for Big Data By Christopher G. Healey

You may also enjoy

K-D-tree

1 minute read

A k-dimensional or k-d tree is a binary tree used to subdivide a collection of records into ranges for k different attributes in each record.

B-tree

2 minute read

In the 1980s B-trees were applied to database management systems, with an emphasis on how to ensure consistent B-tree access patterns to support concurrency ...

Sorting

1 minute read

Sorting is always an important topic in computer science. Theoretically, comparison sorts cannot perform better than O(n lgn) in the best case, better perfor...