453 - Chapter 10
What are the two options to improve I/O to the file system?
- Write an entire file image to the swap space during startup and use demand paging from the swap space - write the file to the swap space when the file is used
Explain working set
- based on locality - replaces process outside the working set/the frame of time that the algorithms choose
What are the benefits of not needing to load an entire program into memory?
- not constrained by size of physical memory - multiple programs run at same time - less i/o needed
What are some complications with NUMA?
- scheduler must track the last CPU each process ran on - threads potentially could be on multiple system boards, and not sure how memory should be accessed in this case
What happens when a page fault occurs?
- trap to the operating system - save the registers and process state - determine that the interrupt was a page fault - check that the page reference was legal, and determinate the location of the page in secondary storage - issue a read from the storage to a free frame: a) wait in q til request is service b) wait for device seek and/or latency time c) begin transfer of the page to a free frame - while waiting, allocate the cpu core to some other process - receive an interrupt from the storage i/o subsystem (i/o completed) - save the registers and process state for the other process (if step 6 is executed) - determine that the interrupt was from the secondary storage device - correct the page table and other tables to show that the desired page is now in memory - wait for the cpu core to be allocated to this process again - restore the registers, process state, and new page table, and then resume the interrupted instruction
What may not be needed in a program's instructins?
- unusual error conditions - extra space for arrays, lists, and tables - some extra options and features
What are the general steps to basic page replacement?
1) find the location fo the desired page on secondary storage 2) find a free frame a) if there is a free frame, use it b) if there is not free frame, use a page replacement algo to select a victim frame c) write the victim to secondary storage, change page and frame tables accordingly 3) read in new frame, change page and frame table 4) continue from where the page fault occurred
What are the two implementations of LRU?
1) have a running clock >> update what is used with the most recent time >> number with the lowest number gets ejected 2) have stack of page number and pull from stack and put on top if used
What are the three major task components of the page-fault service time
1) service the page fault interrupt 2) read in the page 3) restart the process
How many page transfers happen during a page fault and how would one minimize this?
2: page in, page out reduce this by marking dirty bits so only transfer out if written
how does the additional reference bits algo work?
8 bits left most has a one if process used at regular intervals, shift right, if all 0s then not used
What is global replacement?
Allows a process to select a replacement frame from the set of all frames, even if that frame is currently allocated to some other processWhat is local replacement
What is a raw disk?
An array where devices could use large sequential arrays of logical blocks ie databases or data warehouses
Describe the technique of zero fill-on-demand
Before a free frame is filled, it is zeroed out so that the previous information in the frame is erased
What does the FIFO algorithm suffer from?
Belady's anomaly
If the second chance algorithm is implemented with a clock queue, what is its alternative name?
Clock algo
is EAT directly or indirectly proportional to page fault rate?
Directly proportional
What is the formula for effective access time?
EAT = (1 - p) * ma + p * page fault time p is the probability of a page fault (0 <= p <= 1) ma is the memory access time
Is global or local replacement more commonly used and why?
Global because local isolates the process from potential frames that are not commonly used
What example instruction does the textbook state as an example of an instruction that may modify several different locations?
IBM System 360/370 MVC
What is the locality of reference?
If something is referenced in an instruction, the pages around the value will probably be referenced too leading to less page faults despite pre-loading some information
What is the least recently used scheduling algorithm?
If we use the recent past as an approximation of the near future, then we can replace the page that has not been use for the longest period of time
What is the strategy for triggering the page replacement?
Instead of waiting for available frames to hit 0, we do it at a threshold value
What is page fault frequency?
Keep track of the number of faults, and add/remove different chunks of memory based on steady frequency ie. If less than number n (ie 3), then reduce number of frame if greater than number n, then increase number of frames for the process
What is LRU approximation vs LRU?
LRU approximation: don't keep exact order of what was used last (ie additional reference bit, second chance algo) LRU: keep exact order (ie counters and stack)
What are counting base page replacement algos?
Least frequently used most frequently used
What is a solution to limiting the effects of thrashing?
Local replacement algo/priority replacement algo: if one process starts thrashing, it can't steal frames from another process
What is the optimal page replacement algorithm and does it suffer from Belady's anomaly?
Look in the future and replace the page that will not be used for the longest period of time It does not suffer from Belady's
What is a major vs minor page fault?
Minor: it does not have a logical mapping to the page but is in memory. the i.e. process has reference to the shared library in memory, but the process does not have a mapping to it in the page table. OR reclaimed by the free page table but has yet to be zeroed out
Explain the thought process of least frequently used?
Page with smallest count be replaced because if program used more actively == higher ref count
What happens if the number of free frames drop beneath the threshold?
Reaper will more aggressively reap processes, and if that is not enough >> of memory killer will select processes based on the out-of-memory score
What is a disk controller chip?
Rotates the planner and moves the arm Start and stop the motor
What is proportional allocation?
S = sum of virtual memory for all process number of frames to allocate = virtual memory for process / S * available number of frames
what lasts longer, ssd or rotating media disc?
SSD breaks more than magnetic disc
What is the limitation of instructions?
That instructions must fit into physical memory
Where does a process resume from if a page fault occurs?
The exact same place and state but now with the desired page in memory
What is secondary memory?
The memory that holds information not present in main memory
Explain the thought process of most frequently used?
The page with the smallest count was probably just brought in and has yet to be used
Describe the minimum number of frames strategy
The process only allocated a number of the frames as defined by computer architecture
What is a reference string?
The sequence of addresses accessed by the process
What is sparse address space?
Virtual address space that include holes
What are page buffering algorithms?
When a free page pool is used Victim is chose to replace, if need to be written out, use free page pool to write in new value as victim is being written back - also, when the paging device is idle, the modified page is selected and written to the secondary storage
What is local replacement?
When a process can only select from its own set of allocated frames
What is anonymous memory?
When memory in a space may not be of one type (ie space could be associated with a file or not a file)
What is pure demand paging?
When paging is done only as needed
What is copy on write?
When parent and child processes initially share pages unless one of the processes needs to write to the page. Then a page will be made for the process to write to
What is Belady's Anomaly
When the page fault rate increases as the number of allocated frames increases
What is a page fault?
When the process attempts to access a piece of information in the page table/main memory, but it is not loaded
What is thrashing?
When there is not enough frames for multiple processes to run concurrently but these processes are dependent on each other to run
what is a free frame list
a pool of free frames to be used next
Why are NUMA systems potentially better?
accommodate more CPUs and achieve greater levels of throughput and parallelism
What are stack algorithms?
algorithms that can't exhibit Belady's anomaly
What is a frame allocation algorithm?
an algorithm to decide how many frames to allocate to each process
How does the second chance algorithm work?
circular queue has a reference bit which is set to 0 when given second chance look for reference bit that is 0 to replace
What are the components that make up magnetic discs?
disk controller chip arm assembly arm platter spindle
Describe the FIFO replacement algorithm.
first in first out
How does solaris prevent threads from migrating across different domains and incurring memory access penalties?
has locality groups so that CPU in that group can access any memory in the group within a defined latency interval
What is the page fault frequency algo?
have range of ideal page fault rate and add or remove processes to maintain
What is a working set?
if all of local set can be loaded into memory, they we need to make sure one set at a time is loaded and avoid loading more that needed
What are reapers?
kernel routes that are triggered to reclaim pages from all processes in the system (typically excluding the kernel)
What are thrashing solutions?
local replacement algorithms (not an actual solution) Working set Page fault frequency
What is equali allocation?
number of frames available / num processes
what is a locality model?
process will stay using set amount of frames, if not enough allocated, then it will thrash
How does linux prevent threads from migrating across different domains and incurring memory access penalties?
separate free frame list for each NUMA node ensuring that a thread will be allocated memory from the node it is running on
What is the virtual address space?
the virtual view of how a process is stored in memory
What does it mean to be overallocating?
to allocate more virtual memory pages than the physical memory (RAM) available
What is the enhanced second chance algorithm? And what is the order of replacement priority?
to have reference bit and modify bit best to replace to worst to replace (reference, modify) (0, 0) (0, 1) (1, 0) (1, 1)
What is the command the allows copy on write when making a child process?
vfork()
