ML Wiki
Machine Learning Wiki - A collection of ML concepts, algorithms, and resources.

Memory Hierarchy

The Memory Hierarchy

Image

Two main categories:

  • '’volatile’’: forgets what it stores when power goes off
  • '’non-volatile’’: can persist data for a long time

Normally data is moved only between adjacent levels of hierarchy

Cache

  • size: 1 mb or more
  • access time: a few nanoseconds

data and instructions are moved to cache from the main memory when they are needed by the CPU

Main Memory

  • to more from memory to cache/processor: 10-100 nanoseconds

stores all data and instructions

Secondary Storage

  • 10 milliseconds to transfer data from disk to memory

'’Secondary storage’’: disks and other devices that can store large amounts of data

Virtual Memory

'’Virtual Memory’’ is an address space (32 or 64 bits)

  • the OS manages VM keeping needed parts at hand (in the main memory)
  • and the rest on disk

So beware: data can be moved to and fro by the OS| | |### Tertiary Storages

  • very high read/write times
  • may be optical disks stored somewhere

a storage with very large capacity (petabytes, etc)

See also

Sources

  • Database Systems: The Complete Book (2nd edition) by H. Garcia-Molina, J. D. Ullman, and J. Widom