Performance Analysis of Self Limiting Epidemic Forwarding
Self limiting epidemic forwarding is a method of epidemic information dissemination in wireless ad-hoc networks that achieves congestion control by limiting spread (i.e. the number of nodes that receive a given message) and injection rate in order to preserve a meaningful service. We analyze the performance of various methods for spread control: on one hand, the classical method, which consists in decrementing the TTL of a packet when it is transmitted, on the other hand, two methods based on ``aging", where the TTL of a packet may be decremented while it is waiting for transmission in the epidemic buffer. The aging methods are: (selective aging) decrement TTL of a waiting packet when a duplicate is received and (global aging) decrement when any packet is received. The performance metrics are based on injection rates of messages, on amount of redundant information and on spread. We use detailed, realistic simulation for medium scale networks (up to 800 nodes); for networks of any size, we use an analytical method based on fluid approximation and solution of a fixed point problem. We find that the classical method does not perform well. Selective aging improves the performance, and global aging performs much better; it manages to control the spread so that the rate of injection remains good with fixed parameters across a wide range of settings.