A Memory Abstraction: Address Spaces

• The traditional approach of exposing physical memory to processes has several drawbacks.
• As the operating system is by itself a process, another process can trash the operating system by causing inconsistencies in main memory
• Also, multiprogramming is a common trend nowadays and as a result, having processes access physical memory is dangerous.
• A solution is, to use abstract memory, i.e for each process to have a notion of its own address space.
• Each process has its own address space, independent of those belonging to other processes.
• The address perceived by the process is called logical memory address, this is different from actual physical memory location.
• A separate hardware is responsible for converting logical addresses to physical addresses every time the process reads or writes from and to memory.

• Relocation
  
  Every time a process is swapped in or out, it must be correctly located and referenced in order to continue from its previously active state.
  
• Protection
  
  Programs in other processes must be forbidden from referencing memory locations of a given process without valid permissions.
  
• Sharing
  
  Multiple processes accessing same main memory.
  
• Logical Organization
  
  Organize user programs in form of modules so that they can later be broken down into chunks for memory abstraction.
  

There is rarely enough memory to actively hold all code and data structures needed by running processes in memory at any given point of time. The idea is to swap in and out blocks of code as they are required from secondary memory.

Fixed Partitioning
Main memory is divided into static partitions. Any process, given its requirements may be loaded into a partition of equal or greater size.
Note that here processes are not divided into modules, rather memory is divided and entire processes are placed inside a partition.

Partitions may be of two types :

1. Equal Sized
   If program size is beyond that of the partition size, need overlays
   Gives rise to internal fragmentation
2. Unequal Sized
   Fit into smallest possible partition

Dynamic Partitioning

Here, partitions are created dynamically as per the requirement/size of the process in question.

• Partition created to be exact fit for asking process size. No Internal fragmentation.
• When processes are pulled out, it creates holes in memory space. This causes external fragmentation.
• Need to perform a process known as compaction to regularly shift partitions and remove holes created by processes leaving main memory. This is a costly process and cannot be done repeatedly.

• Main memory is divided into equal sized partitions. Each partition is called a frame.
• Processes are divided into equal sized chunks where each chunk is called a page
• To load a process, all of its pages are loaded into frames which need not be contagious. The OS takes care of the order of execution.
• Operating System maintains a page table for each process which contains the frame location for each page.
• Paging causes zero internal fragmentation and is optimal among static partitioning techniques.
• Paging causes minimal external fragmentation.

1. The key difference between fixed and dynamic partitioning is
1. fixed partition sizes are fixed while dynamic partition sizes are determined at runtime depending on size of process
2. fixed paritioning is done once while dynamic partitioning is done multiple times
3. fixed partitoning involves disk while dynamic partitioning involves memory
4. None of the above
2. An executing process perceives
1. physical address
2. logical memory address
3. disk address
4. None of the above
3. Which of the following causes internal fragmentation
1. swapping(dynamic partitioning)
2. swapping(fixed partitioning)
3. segmentation
4. paging