Storage hierarchy

from Wikipedia, the free encyclopedia
Diagram of the storage pyramid (in English )

In computer science , memory hierarchy describes the arrangement of memories in a computer architecture from the point of view of the main processor , arranged according to decreasing access speed , decreasing costs , increasing storage capacity and increasing access unit. The graphic representation of this hierarchy takes place i. d. Usually in a triangular shape and is called a storage pyramid.

more details

A (utopian) ideal memory would be extremely inexpensive, infinitely large, would require no access time and would not lose its content even if there was no power ( non-volatile memory ). If such a memory were available, memory management would be trivial. In reality, however, fast memories are orders of magnitude more expensive than slower memories, and they are usually volatile. For this reason, an attempt is made to find a compromise between speed, costs and persistence by creating the memory in a memory hierarchy and trying to use the advantages of the various components through clever memory management while at the same time circumventing their disadvantages.

This results in the shape of a storage pyramid: At the top there is very little faster and more expensive storage, while at the bottom there is much slower, very inexpensive storage. So while the speed decreases from top to bottom, the storage capacity grows to about the same extent.

A storage hierarchy used on many computers today is:

  1. Processor register , order of magnitude: byte-kilobyte, access time i. d. Usually with processor clock (just as fast as CPU)
  2. Processor caches , size: kilobyte-megabyte, access time: around a few dozen clock cycles (depending on the level )
  3. Main memory , size: megabyte – gigabyte, access time: hundreds of clock cycles (today almost exclusively DRAM , e.g. EDO-RAM , SDRAM , DDR-SDRAM ), access unit cacheline
  4. Distributed memory , size: gigabyte – terabyte, access time: network latency
  5. Mass storage , order of magnitude: gigabyte – terabyte, access time: millions of clock cycles (e.g. hard disk , solid state drive ), access unit block
  6. Removable storage media , size: gigabyte – terabyte, access time: milliseconds to several minutes, but cheaper than mass storage (e.g. DVD , CD , USB stick , floppy disks , magnetic tape )

In addition to the high costs, another reason why very fast memory can only be used to a limited extent is the following: The access speed and storage capacity are also influenced in the higher levels by the spatial arrangement. The main processor is space-limited and there is a relatively high cooling requirement. The vertical arrangement of the storage groups (i.e. the utilization of the third dimension) can therefore only take place within narrow limits.

For better organization of the memory (e.g. optimization of the access time for frequently required data) memory contents are often transparently moved within the hierarchy, e.g. B. parts of the hard disk are loaded into the data carrier cache , on the other hand parts of the main memory are swapped out to the swap file ( virtual memory ).

Databases

The different storage levels of a database are sometimes referred to as primary storage , secondary storage or tertiary storage depending on their speed . The primary storage is usually the main storage, while secondary storage means fast permanent storage and tertiary storage means slow archiving medium.

Memory management

Hierarchical storage management systems automatically move data that has not been required for a long time to storage at a lower hierarchical level. Also, information lifecycle management automates such processes.

Individual evidence

  1. Storage technologies at a glance: In the delta of the data flow - Linux-Magazin , edition 11/2004
  2. Andrew S. Tanenbaum ; Modern Operating Systems, Second Edition (English); 2001, ISBN 0-13-092641-8 , p. 189; see also Modern Operating Systems (International Edition) (2nd Edition) , at GetTextbooks.com