OS Distributed Shared Memory

Remote procedure call has its attraction, but many computer programmers still prefer to use shared memory, even on multicomputer system.

It is also possible to preserve the illusion of shared memory reasonably well, even when it doesn't actually exist just by using a computer technique, that is called as distributed shared memory (DSM).

With distributed shared memory, each page is located in one of memories. And each machine has its own virtual memory and page tables.

Here, when a central processing unit (CPU) does a LOAD and STORE on a page, then it doesn't have, a trap to OS occurs. The OS then locates the page and ask the central processing unit currently holding it just to unmap the page and send it over interconnection network.

Now, when it arrives, then the page is mapped in and the faulting instruction is restarted.

In effect, the OS is just satisfying page faults from remote random access memory (RAM) instead of from the local disk.

And therefore, to the computer user, the machine looks as if it has shared memory.