That 'how stuff works' article is a bit dated...
Virtual Memory is just a way of giving every process on a computer the same amount of space to play in. Actual physical memory used is shared out to the processes on demand. If that demand becomes greater than the physical memory in total then the disk is used for least recently used pages of RAM to be written out to. In most OS's its desirable for the OS to keep a proportion of RAM unsued to allow fast access to new demand, rather than wait for the eviction of a page as described above.
When a process leaks knowledge of the memory it's using it stops referring to it and often continues to demand new pages. This eventually ends up on disk as an evicted least recently used page. Then once that resource is exhausted you see in windows warnings that there's low VM available. Adding more page file usually just delays the onset of this again. Finding the leaky program in the taskmanager or windbg is the best fix.
You really can't turn off VM in most OS's but you can change it's behaviour a bit. The underlying hardware really dictates this, as the address space for all 32/64 bit os's is virtual just like the processes.