operating systems - memory, files

8/8/2019 Operating Systems - Memory, Files

1/46

Memory ManagementRobert Dickerson


2/46

Why is it generally inefficient to allow onlyone process to be in memory at a time?


3/46

If the single process blocks for I/O, noother processes can use the processor

Answer


4/46

What would happen if a system allowedmany processes to be placed in mainmemory, but did not divide memory into

partitions?


5/46

The processes would all share their

memory

Any malfunctioning or malicious processcould damage any or all of the other

processes

Answer


6/46

Name the two types offetch strategies

Describe when one might be moreappropriate than the other


7/46

Demand fetch andanticipatory fetch

If the system cannot predict future memoryusage with accuracy, then the loweroverhead of demand fetching results in

higher performance and utilization

If program exhibits predictable behavior,anticipatory fetch strategy is best


8/46

When is noncontiguous preferable tocontiguous memory allocation?

Contiguous vs.

Noncontiguous Memory


9/46

Answer

Noncontiguous is better when availablememory contains no area large enough to

hold the incoming program in a contiguous

piece


10/46

What sort of overhead is involved in a

noncontiguous memory allocation

scheme?


11/46

You need to store:

available blocks

blocks that belong to separate processes

where those blocks reside in memory


12/46

Describe the benefits and drawbacks oflarge and small partition sizes


13/46

Larger partitions allow large programsto run, but result in internal fragmentation

for small programs

Small partitions reduce the amount ofinternal fragmentation and increase the

level of multiprogramming by allowing more

programs to reside in memory at once


14/46

Explain the difference between internaland external fragmentation


15/46

Internal fragmentation occurs infixed-partition environments when a

process is allocated more space that it

needs.

External fragmentation occurs invariable-partition environments when

memory is wasted to holes developing inmemory between partitions.


16/46

Why are segmentation/pagingsystems appealing?


17/46

Segmentation/paging offers architectural

simplicity of paging and access control

capabilities of segmentation


18/46

Why are large page sizes more favorable in

todays systems than they were decades

ago?


19/46

Cost of memory is cheaper, applicationsmore memory intensive

Cost of internal fragmentation is less of aconcern

Large pages require the system to perform

fewer costly I/O operations


20/46

What restriction does assembly time

or load time binding of address place

on the swapping of a process?


21/46

If addresses are not bound at runtime, thenthe process must be swapped into thesame memory location from which it was

swapped out


22/46


23/46

Compaction - if addresses can be boundat runtime, we can move the used memoryaround so that the unused memory is

contiguous and reset the base register.

Break a process into 2 or more smallersections to increase likelihood small

memory spaces will be used

Loosen the requirement that processmemory be contiguous. Solution- paging


24/46

What is a TLB?


25/46

Translation Look-aside Buffer hardware method of speeding up

translation of logical to physical memory

addresses

acts as a cache for more recent translations

page number is compared to all page-framepairs in parallel - very fast!


26/46

What is involved with a context switch?


27/46

Occurs when a process is interrupted andanother process gains control of the

system

Save registers, program counter, page tableor segment pointers, stack, etc.


28/46


29/46

compiled into separate units(code, data, stack)

user specifies a segment number and offsetinstead of an address.

Segments can have variable lengths

requires a field in segment tablecontaining the limits of the offset value


30/46

What property must code have in able tobe shared?


31/46

It must be reenterant

A computer program or routine is described asreentrant if it can be safely called recursively andconcurrently from multiple processes.

To be reentrant, a function must:

hold no static data

must not return a pointer to static data, must work only on the data provided to it by

the caller, and must not call non-reentrant

functions.
http://en.wikipedia.org/wiki/Concurrency_%28computer_science%29http://en.wikipedia.org/wiki/Computer_programhttp://en.wikipedia.org/wiki/Subroutinehttp://en.wikipedia.org/wiki/Concurrency_%28computer_science%29http://en.wikipedia.org/wiki/Concurrency_%28computer_science%29http://en.wikipedia.org/wiki/Recursion_%28computer_science%29http://en.wikipedia.org/wiki/Recursion_%28computer_science%29http://en.wikipedia.org/wiki/Subroutinehttp://en.wikipedia.org/wiki/Subroutinehttp://en.wikipedia.org/wiki/Computer_programhttp://en.wikipedia.org/wiki/Computer_program


32/46

What is virtual memory?


33/46

Allows execution of processes that may notbe completely in memory.

Allows programmers to think of memoryas a large uniform space


34/46

Why is virtual memory almost never usedin a realtime system?


35/46

Dependability of the system

Predictable in time required to complete

operations

Virtual memory creates less predictabledelays in memory-related operations


36/46

What happens on a page fault?


37/46

Page table for the process is used todetermine if the page corresponding tological address is invalid

If it is, OS chooses a free frame or selects apage to be paged out, and gets the pagefrom backing store

Page table is updated to reflect that page isnow in memory and instruction is restarted


38/46

What is Beladys Anomoly?


39/46

Beladys anomaly states it is possible tohave more page faults when increasing the

number of page frames while using FIFO

method of frame management.

Previously, it was believed that an increasein the number of page frames would always

provide the same or fewer page faults


40/46

example

3 2 1 0 3 2 4 3 2 1 0 4

3 page frames


41/46

Why is it that no page replacementalgorithm can be optimal?


42/46

Optimal algorithms would be where thepage that will not be used for the longest

time will be swapped out

No algorithm can be optimal because it isimpossible to foresee the future page usage


43/46

How do you implement aLRU page replacement?


44/46


45/46

What is an inverted page table?


46/46

Instead of mapping virtual page to a frame,map the frames to the pages

Results in a table whose number of entriescorresponds to the number of frames

operating systems - memory, files

Documents