A low overhead dynamic memory management system for constrained memory embedded systems

Embedded systems programming often involve choosing the worst case static memory allocation for most applications over a dynamic allocation approach. Such a design decision is rightly justified in terms of reliability, security and real time performance requirements from such low end systems. However with the introduction of public key cryptography and dynamic reconfiguration in IP enabled sensing devices for use in several “Internet of Things” applications, dynamic memory allocation in embedded devices, is becoming more important than ever before. While several embedded operating systems like MantisOS, SOS and Contiki provide dynamic memory allocation support, they usually lack flexibility or have relatively large memory overhead. In this paper we introduce two novel dynamic memory allocation schemes, ST_MEMMGR (without memory compaction) and ST_COMPACT_MEMMGR (with memory compaction), with a close compliance with the libc memory allocation API. Both designs take into account the very limited RAM (1KB - 64KB) in most microcontrollers. Experimental results show that ST_MEMMGR has a 256 - 5376 bytes lesser memory overhead than similar non-compaction based open source allocators like heapLib and memmgr. Similarly, ST_COMPACT_MEMMGR is observed to have 33% smaller memory descriptor as compared to Contiki´s managed memory allocator with similar performance in terms of execution speed.