Memory Management

Question 1

What is paged-based memory management?

Answer 1

Virtual memory is divided into pages, physical memory is divided into page frames (same size as pages). Pages tables map pages to frames

Question 2

What is segment based memory management?

Answer 2

Virtual memory mapped into variable-length segments. Uses segment registers supported by hardware

Question 3

What are the ways that hardware supports memory management?

Answer 3

MMU, registers, cache, memory translation

Question 4

What is the MMU?

Answer 4

Memory management unit - included in the CPU and does address translation VA->PA. Reports faults if access is not allowed

Question 5

What is the TLB?

Answer 5

Translate Lookahead Buffer. Part of the hardware cache that stores VA->PA mappings

Question 6

What is allocation on first touch (aka demand paging)?

Answer 6

PA is only allocated when the program first tries to access it.

Question 7

What is the purpose of a validity bit in a page table?

Answer 7

1 = page is valid and in memory, 0 otherwise ==> MMU will raise a fault, OS must decide what to do (swap, allocate PA, or page fault)

Question 8

How large is a page table?

Answer 8

Total virtual addresses = 2^32
number of virtual pages = 2^32 / page size
need total / pages * entry size ==> does not scale

Question 9

Explain a multi-level page table

Answer 9

Consists of an outer-page table and inner page table. First part of address is an index into outer table, which points into inner page table (which is a real page table). Then that points to the page frame. Size of inner page table == page frame. Size of outer page table == 2^remaining bytes

Question 10

What are the benefits of a multi-level page table?

Answer 10

You do not have to allocate a page table for the entire page space. Any gaps in memory equal to page size * 2, can result in NULL in the outer page table

Question 11

What are the pros/cons of adding more levels to page tables?

Answer 11

More levels == smaller page size == less wasted memory, but also means slower lookup times

Question 12

What is an inverted page table?

Answer 12

IDK

Question 13

What is external fragmentation?

Answer 13

When memory page allocation gets fragmented so that there is enough free memory to fulfill an alloc request, but it is not contiguous (and therefore the memory cannot be allocated)

Question 14

What is the memory allocator?

Answer 14

Determines the VA->PA mappings

Question 15

What is the buddy allocator?

Answer 15

Start with memory area of 2^x, when a new request comes in, divide area by 2 until it is just big enough

Question 16

What is the slab allocator?

Answer 16

Creates caches for common objects / sizes, sits on top of physical memory. Avoids fragmentation

Question 17

What is demand paging?

Answer 17

When a page has been swapped out of main memory to disk and needs to be put back into memory because of an access –> hardware will trap to OS, which will swap the page before returning control back to the process

Question 18

What is a pinned page?

Answer 18

A page that cannot be swapped

Question 19

What is page replacement?

Answer 19

When you swap pages to disk. Want to swap pages that won’t be used soon - LRU or pages that haven’t been modified since reading them from disk (dirty bit in MMU)

Question 20

When are pages swapped out?

Answer 20

When amount of memory in main memory hits a certain threshold and when CPU usage is low

Question 21

What is Copy-on-Write?

Answer 21

When fork a process, a lot of the memory is the same. Set P2’s pages to read only and just point them to P1’s PA. When P2 wants to modify one, then you copy the page over (via a trap)