Memory Management(1)

Memory management (1)

• Memory Hierarchies
  • Computers typically have memory hierarchies
    • Registers, L1/L2/L3 cache
    • Main memory
    • Disks
  • “Higher memory” is faster, more expensive and volatile, “lower memory” is slower, cheaper, and non-volatile
  • Memory can be seen as one linear array of bytes/words
  • The operating system provides a memory abstraction

• OS responsibility
  • Allocate/deallocate memory when requested by processes, keep track of used/unused memory
  • Transparently move data from memory to disc and vice versa
  • Distribute memory between processes and simulate an “infinitely large” memory space
  • Control access when multiprogramming is applied

Models

• Approches: Contiguous vs. Non-Contiguous (L13 - p7)
  • Contiguous memory management models allocate memory in one single block without any holes or gaps
  • Non-contiguous memory management models are capable of allocating memory in multiple blocks, or segments, which may be placed anywhere in physical memory

Partitioning

• Contiguous approaches
  • Mono-programming: one single partition for user processes
  • Muti-programming with fixed partitions
    • Fixed equal sized partitions
    • Fixed non-equal partitions
  • Muti-programming with dynamic partitions

• Mono-programming - No Memory Abstraction
  • Only one single user process is in memory/executed at any point in time (no multi-programming)
  • A fixed region of memory is allocated to the OS/kernel, the remaining memory is reserved for a single process (MS-DOS worked this way)
  • This process has direct access to physical memory (i.e. no address translation takes place)
  • Properties:
    • Every process is allocated contiguous block of memory, i.e. it contains no “holes” or “gaps” (, non-contiguous allocation)
    • One process is allocated the entire memory space, and the process is always located in the same address space
    • No protection between different user processes required (one one process)
    • Overlays enable the programmer to use more memory than available (burden on programmer)
  • Shortcomings of mono-programming:
    • Since a process has direct access to the physical memory, it may have access to OS memory
    • The operating system can be seen as a process - so we have two processes anyway
    • Low utilisation of hardware resources (CPU, I/O devices, etc.)
    • Mono-programming is unacceptable as multiprogramming is expected on modern machines
  • Direct memory access and mono-programming is common in basic embedded systems and modern consumer electronics, e.g. washing machines, microwaves, car’s ECUs, etc.
  • Simulating Muti-Programming
    • Simulate multi-programming through swapping
      • Swap process out to the disc and load a new one (context switches would become time consuming)
      • Apply threads within the same process (limited to one process)

• Muti-programming
  • A Probabilities Model (L13 - p15)
    • There are n processes in memory
    • A process spends p percent of its time waiting for I/O
    • CPU Utilisation is calculated as 1 minus the time that all processes are waiting for I/O: e.g., p = 0.9 then CPU utilisation = 1 - 0.9 => 0.1 (1 - p)
    • The probability that all n processes are waiting for I/O (i.e., the CPU is idle) is pn, i.e. 
    • The CPU utilisation is given by 
    • CPU utilisation goes up with the number of processes and down for increasing levels of I/O

Partitioning

• Fixed Partitions
  • Divide memory into static, contiguous and equal sized partitions that have a fixed size and fixed location
    • advantages:
      • Any process can take up any (large enough) partition
      • Very little overhead and simple implementation
      • The operating system keeps a track of which partitions are being used and which are free
    • Disadvantages:
      • Low memory utilisation and internal fragmentation: partition may be unnecessarily large
      • Overlays must be used if a program does not fit into a partition (burden on programmer)
  • Divide memory into static and non-equal sized partitions that have a fixed size and fixed location
    • Reduces internal fragmentation
    • The allocation or processes to partitions must be carefully considered
• Allocation method
  • One private queue per partition:
    • Assigns each process to the smallest partition that it would fit in
    • Reduces internal fragmentation
    • Can reduce memory utilisation (e.g., lots of small jobs result in unused large partitions) and result in starvation
  • A single shared queue for all partitions can allocate small processes to large partitions but results in increased internal fragmentation