Dynamic Select Approach for Memory Allocation

Abstract

When we need to use Memory allocation for relatively huge datasets, then we may have a possibility to encounter the exception that is OutOfMemoryException. This exception shows that memory is not available for the allocation. But exception does not occur due to limited memory system, it usually occurs due to non availability of virtual address space for that byte of data. This issue is because of the current implementation of memory allocation which uses single array byte as backing store. When the data set is huge the backing store of memory allocation space also requires more contiguous memory than that is available in the virtual address space. If there is no contiguous memory available for the process then it encounters the exception of OutOfMemoryException even there is enough space available but not continuous. This research proposed an approach for dynamically selecting the best memory allocator for every application. The proposed approach does not need any type of contiguous memory for storing the data in stream. This approach uses a dynamic list of small chunks as backing storage that are allocated on demand when the stream is used. If there is no contiguous memory available in the Stream then memory allocation can be done from these small chunks of memory with no OutOfMemoryException.