Thrashing

Question 1

what is thrashing? what causes it?

Answer 1

high paging to disk activity, process spends more time paging the disk than executing - caused by not having enough frames allocated

caused by too much multiprogramming, each process getting an allocation that is too small

Question 2

how do we limit thrashing?

Answer 2

using a local replacement algorithm

• page fault evicts a page of the same process
• one process thrashing cannot cause others to thrash by stealing from others

or provide the process with the number of frames it needs

Question 3

how do we know how many frames a process needs?

Answer 3

define a locality model

• a model of process execution
• process locality is a set of pages actively used together
• proccess moves from locality to locality as it executes

locality of a function is
- instruction
- local variables
- subset of global variables
we leave the locality on return from function

Question 4

how do we use locality?

Answer 4

locality determined by program structure and data structures

allocate enough frames to hold locality

• intially pages in locality are faulted
• no more faulting until locality changes

thrashing occurs when frame allocation is too small to hold locality

Question 5

what is the working set?

Answer 5

• based on the assumption of locality, the paramater Δ defines the working-set window
• we examine the most recent Δ page references, they are the working set
• if a page is in use, it will be in the set
• if a page is no longer in use, it will be dropped from the set Δ time units after its last reference
• the size may change over time

the demand D is
D = sumof WSSi
where WSS is working set size
thrashing occurs if D is greater than the number of frames