Obtaining memory available to JVM at runtime

Question

I\'m trying to sort a bunch of data such that that the size of data input to the program can be larger than the memory available to the JVM, and handling that requires exter

Answer 1

Using the methods of Runtime, as the others suggested, is fine, as long as you take some things into consideration:

1) freeMemory() is a lower bound on the actual available memory, because memory that is unreferenced and ready for GC is considered as used. Running System.gc() before the call may return a more accurate result.

2) totalMemory() can change - it only indicates the current total heap size, and the heap can expand/shrink by the JVM during runtime, depending on its usage. You can use maxMemory() to get the actual maximum.

Answer 2

You can use the Runtime class to get the amount of memory available.

Runtime r = Runtime.getRuntime();
System.out.println(r.totalMemory());

There are various other memory details you can get from the Runtime object - see Runtime class.

Answer 3

Check out these methods on the java.lang.Runtime class:

freeMemory

totalMemory

maxMemory

Example

Runtime rt = Runtime.getRuntime();
System.err.println(String.format("Free: %d bytes, Total: %d bytes, Max: %d bytes",
  rt.freeMemory(), rt.totalMemory(), rt.maxMemory()));

Also note that if the total amount of memory is being exhausted you can always start the JVM with a larger amount of heap allocated using the -Xmx JVM argument; e.g.

java -Xmx256M MyClass

Answer 4

In theory yes, using Runtime.getRuntime().maxMemory().

In practice, there are some problem you need to address:

1. You need to figure out how many application objects are going to fit in a given number of bytes of memory. AFAIK, there is no simple / efficient way to do this within a running application.

2. You don't want to try to use all available heap space. If you push your percentage heap residency too high, you risk making the GC horribly inefficient.

3. The maxMemory() method only tells you how big the heap in virtual memory. The physical size can also be a factor (especially if physical size << virtual size), and there's no portable way to figure that out.

If I was trying to implement this application, I think I'd probably just make the in-memory sort size a configuration parameter or command-line option.