22.1.9. Thrashing

If a process routinely accesses more virtual memory than it has physical memory due to insufficient physical memory it suffers thrashing as we saw.

Thrashing happens when a program needs more memory than what is physically available. This results in the operating system constantly swapping data between the disk and RAM. Instead of running effectively, the system gets bogged down because it is busy moving data around. When this occurs, the CPU spends more time managing memory access rather than executing the program's instructions.

Imagine trying to cook dinner in a small kitchen full of clutter. Every time you need an ingredient, you must first clear space on the counter, which takes up time. Similarly, thrashing occurs when the system is too cluttered with too many processes needing access to memory.

The set of popular pages corresponding to a program at a given time is called its working set. The pages the set of pages that that a particular program has accessed in the recent past is there in the working set.

The 'working set' refers to the pages a program needs to keep in memory to execute efficiently. If all necessary pages fit into the physical memory, the program runs smoothly. However, if these pages exceed the available physical memory, thrashing starts since not all pages can be accessed quickly by the program.

Think of the working set like the specific tools a carpenter needs for a project. If he has all his tools on hand, he works efficiently. But if he has to keep running to the toolbox to find tools that don’t fit on his workbench, he becomes inefficient, similar to how a program becomes slow due to thrashing.

To handle this situation, we can either have more we can allocate more physical memory, to be made available to this process. So, I need to handle thrashing to reduce access, because I don’t have all pages in the working set in main memory. How can I improve the situation? I will have to increase memory that is allocated to this program. The program must be temporarily suspended in the interest of the rest of the system.

One way to alleviate thrashing is by increasing the physical memory allocated to a program, allowing it to accommodate its working set. If this is not possible, the operating system may choose to temporarily suspend the program causing thrashing, freeing up resources for other processes. This allows other programs to run smoothly without the performance hit caused by excessive page swapping.

This situation is like a student needing more desk space for their projects. If the desk is too small, the student can either get a bigger desk or temporarily pack up some projects to allow for easier access. By managing space, they can continue working without delays.

Another way to reduce thrashing is to have better algorithms and data structures for the program so that the locality of this program can be improved. When the locality of the program is improved, effectively the working set size of the program, the number of distinct pages in the working set, reduces. When this happens, the working set size therefore reduces.

Improved program locality means that a program can access the required data and instructions more frequently without needing to swap pages. By using better algorithms that enhance how data is accessed—keeping related data close together—programs can operate effectively with limited memory resources. This reduces the number of distinct pages needed and minimizes chances of thrashing.

Imagine a librarian organizing books in a way that similar genres are placed next to each other. This organization allows patrons to find what they need quickly without searching the entire library, just like improved locality helps programs find data faster without accessing different pages.

--