Epsilon doesn't clear the heap memory of unused objects; it only allocates memory. When the JVM runs out of memory, it shuts down with an OutOfMemoryError error. If a heap dump is enabled, Epsilon will perform a head dump, after throwing an OutOfMemoryError error.

Epsilon GC uses a simple, lock-free Thread Local Allocation Buffer (TLAB) allocation. This works by allocating contiguous sections of memory in a linear manner. One of the main benefits of TLAB allocation is that it binds a process to use the memory that has been allocated to it.

The barrier set used by Epsilon GC is completely empty, hence the name, no-op. Since it doesn't reclaim memory, it doesn't need to bother itself with maintaining the object graph, object marking, object compaction, or object copy.

Will you ever have any latency overhead with Epsilon? Yes, it is possible. Your application might experience latency overhead while Epsilon allocates memory—that is, if the memory size is too large and the chunk of memory that is being allocated is also large.