Augmenting the Repertoire of Design Patterns for Self-Organized Software by Reverse Engineering a Bio-Inspired P2P System

Abstract

Investigations of self-organizing mechanisms, often inspired by phenomena in natural or societal systems, have yielded a wealth of techniques for the self-adaptation of complex, large- and ultra-large-scale software systems. The principled design of self-adaptive software using principles of self-organization remains challenging. Several studies have approached this problem by proposing design patterns for self-organization. In this paper, we present the results of applying a catalog of biologically inspired design patterns to Mycoload, a self-organizing system for clustering and load balancing in decentralized service networks. We reverse-engineered Mycoload, obtaining a design that isolates instances of several patterns. This exercise allowed us to identify additional reusable self-organization mechanisms, which we have also abstracted out as design patterns: SPECIALIZATION, which we present here for the first time, and a generalized form of COLLECTIVE SORT. The pattern-based design also led to a better understanding of the relationships among the multiple self-organizing mechanisms that together determine the emegent dynamics of Mycoload.