This is something for the operating system administrator to check. You can not check this from within Oracle. You might have 99.9% cache hits, but this does not tell you if the cache hits were in memory or on disk? If you do not have enough RAM, then your SGA will be paged or swapped to disk, and cache hits will occur on disk. This is why you cannot check from within Oracle.
On Windows, launch Task Manager (Ctl-Shift-Esc) and look at the "Performance" tab, and specifically in the box labeled "Physical Memory". In this box, find the line "Total" and write down the number. Now, look at the bottom right corner of Task Manager where it says "Mem Usage". There are two numbers, write down the first number only. (Mem usage has a second value that includes your RAM PLUS pagefile, so ignore that number for now.) The second number you wrote down should be 50% of the first number you wrote down. If it is higher, then add more memory. Otherwise, anytime the memory usage is >50% Windows will page more agressively and your SGA performance will suffer.
On Unix, the steps are different, but you can use a bit more memory. You want to keep the amount of memory used < 60% RAM.
On Windows 2000 Server, you can install up to 4 GB RAM. If your operating system and application processes together use more than 2 GB RAM, then you will find Windows begins to page blocks from RAM to disk more agressively, even if no one is using your server.
It's a good rule of thumb to install as much RAM as you are able, simply because it's the fastest storage-device out there. If RAM is available, the operating system will use it. Generous space allocations can be given to things like Oracle.
The thing to watch is not "how much RAM is being used," because the OS will use as much RAM as it can. So, for that matter, will Oracle, even though its memory is virtual.
Instead, Watch for delays. Watch for situations where the Oracle daemon is being put into wait-states due to operating system page-I/O waits: "thrashing." Watch for resource-shortages that are actually forcing things to slow down. "It's only a problem when it presents a problem."