File-as-heap

File backed Heap object store, primarily targetted for Caching applications.

Issues in traditional memory approaches

  • Fragmentation occurs due to un-even memory requirements.
  • Caches are always symmetric in nature and every cache object are of same shape and size. Hence memory fragmentation can be minimized in this.
  • Famous memory allocators like libumem, hoard etc do not provide, file backing.
  • Caches are meant to restore the state, hence maintaining off-line history is an important requirement.
  • Even in case of file backing, deleted / timed-out sessions are not really deleted from the file.
  • Hence over the perid of time, file size grow to enormous extent and we are forced to go for new file.

Features of File-as-heap

  • File backed heap. Restoration from the heap binary file is possible during init stage.
  • Every object will be of same size. The total size of the file will be automatically word aligned.
  • MMAP policy is used. Hence the File I/O bottleneck does not happen. This improves performance.
  • The custom heap is aligned in buckets. Hence every bucket can be re-used, re-cycled. The file size of the heap file never increases. Still the recycle is achieved.
  • Easy integration possible. The allocate / deallocate / walkHeap provides shares only the pointers. Hence the meta-data involving in this product is merely pointers. nothing else.

Features planned for future versions

  • HTTP based statistics reporting.
  • Shared library based packaging.
  • COM object. Reusable.

'''