What is thrashing? Why does it occur?

Question

In operating systems that implement a virtual memory space the programs allocate memory from an address space that may be much larger than the actual amount of RAM the system possesses. The OS is responsible for deciding which programs "memory" is in actual RAM. It needs a place to keep things while they are "out". This is what is called "swap space", as the OS is swapping things in and out as needed. When this swapping activity is occurring such that it is the major consumer of the CPU time, then you are effectively thrashing. You prevent it by running fewer programs, writing programs that use memory more efficiently, adding RAM to the system, or maybe even by increasing the swap size.

A page fault occurs when the memory access requested (from the virtual address space) does not map to something that is in RAM. A page must then be sent from RAM to swap, so that the requested new page can be brought from swap to RAM. As you might imagine, 2 disk I/Os for a RAM read tends to be pretty poor performance.

Thrashing It is a state in which our CPU perform 'productive' work less and 'swapping' more. CPU is busy in swapping pages, so much that it can not respond to user program as much as required. Why it occurs In our system Thrashing occurs when there are too much pages in our memory, and each page referes t an other page. The real memory shortens in capacity to have all the pages in it, so it uses 'virtual memory'. When each page in execution demands that page that is not currently in real memory (RAM) it place some pages on virtual memory and adjust the required page on RAM. If CPU is so much busy in doing this task, thrashing occurs.

I know this question has been asked long ago, but I just wanted to share information with others.

The term thrashing is actually related to the virtual memory, that an operating system uses in order to provide extra amount of memory or space for the processes. What dose it actually mean by the term thrashing is that, when the process is ready to be loaded in the memory, only a few or some pages(parts) of the process are loaded in the actual physical memory, and rest are in the swap-space(the virtual memory or the disk).

Now if the page that the process needs to execute, is not loaded in the memory, it generates a page-fault and asks the OS to replace the page. Here the process resumes its execution.

Some times, the page replaced by the OS is again required by the process therefore it again asks the OS to load it in the memory, replacing some other page and so on. since the process is not executing, therefore the CPU utilization is 0, However the disk read and write are at the peak.

Our OSs are designed in such a way that when the CPU utilization decreases it initiates another process in the memory. The next process have to wait now because the first process is busy. Again since the CPU is not being utilized or it is 0(in our example), the OS initiates another process, and the same thing happen.

Therefore, the CPU utilization decreases to an extreme minimum level, while the processes are busy reading and writing(swapping the pages). This is called thrashing!

Logical addresses are generated by CPU which are in fact not real memory location but a process thinks that it is the actual memory location. A complete process is divided into different parts which are stored in pages in logical memory but only sub parts or only some pages are allocated actual memory for execution which are required at that point of time, whereas other pages reside in logical memory and are not allocated a physical address. Now if any other pages are required to load and there is not enough frame to allocate it to assign, an interrupt occurs which causes a page fault and a replacement algorithm is needed to remove some pages from frames and load the current page required.

Now suppose there are not enough frames to meet requirements for all the pages needed to be loaded. Then the frame removed will again demand to be loaded and again page fault will occur. This goes in loop. Meanwhile, since the process is not able to execute, CPU utilization will be low, and the dispatcher will load more processes in the CPU which degrades the scenario even more.

This process is known as thrashing. Thrashing generally occurs when enough frames are not allocated to a process than needed.

Ways to prevent thrashing:

1. instruct mid-term scheduler to swap out some of the process too recover from thrashing
2. instructing the dispatcher not to load more processes after a threshold