A membrane computing inspired packing solution and its application to service center workload distribution Daniel Moldovan, Georgiana Copil, Ioan Salomie, Ionut Anghel, Tudor Cioara Technical University of Cluj-Napoca, Cluj Napoca,Romania {daniel.moldovan,georgiana.copil,ioan.salomie,ionut.anghel,tudor.cioara}@cs.utcluj.ro Abstract—This paper presents a bio-inspired system based on membrane computing for solving packing problems in complex dynamic systems where the characteristics of the packed items change continuously. For representing such systems, a bio- inspired model is defined, having as core entities cells and molecules, the packing problem translating into the problem of matching molecules to cells. A symbiotic relationship involving a mutual exchange of chemicals and energy between cells and molecules is defined and used to control the matching process. The system is evaluated in the context of workload distribution in service centers, having as goal the reduction of service center energy consumption by minimizing the number of used servers without affecting the workload resource requirements. Keywords-bio-inspired, membrane computing, cloud comput- ing, workload distribution, energy efficiency I. I NTRODUCTION Membrane computing is a branch of bio-inspired computing which extracts computing models from the architecture and the functioning of living cells [1] in order to design so called P-systems. The expressiveness of membrane inspired models has been studied in [2], which shows that even systems which lack features like polarization, label change or division of non- elementary membranes describe universal Turing machines. The computational power of different variants of P-systems has been studied in [3] which targets P-systems with active membranes and two polarizations per membrane and in [4], which focuses on P-systems with mobile membranes, showing that P-systems are computationally efficient and equivalent with Turing machines, being able to solve NP-problems in polynomial time under certain conditions. Over the last years the energy efficiency management of ser- vice centers has emerged as a critical environmental challenge. A U.S. Environmental Protection Agency report to Congress [5] describes an alarming trend in the rise of electricity con- sumed by service centers and their additional infrastructure. One of the major sources of the energy consumption problem is the inefficient utilization of computing resources. According to [6],in a service center about 30% of servers having an average utilization ratio between 5 and 10 percent. This under- utilization provides a huge opportunity for organizations to reduce the service center energy consumption by employing energy-aware workload distribution techniques to reduce the workload dispersion and turning off unused servers. In this paper we present a generic membrane-computing inspired computational model which can be applied for solv- ing packing problems. We consider as packing problem any problem in which items need to be grouped with respect to some constraints, from distributing virtual machines in cloud computing infrastructures to packaging different items into boxes for shipment from warehouses to stores. The defined membrane-computing inspired model extracts rules from the biological cell behavior and symbiotic relationships found in nature and applies them in building a rule-based packing solution. For validating the described approach, the presented computational model is applied to a service center workload distribution scenario and evaluated in terms of decision time and solution quality against a best fit first approach. The rest of the paper is structured as follows: Section II presents the state of the art and real-world applications for membrane-inspired systems, Section III introduces the membrane-inspired context representation model, Section IV describes the symbiotic process between cells and molecules, Section V details the process through which the system evolves, Section VI presents two evaluation scenarios for the membrane-inspired system, and Section VII concludes the paper. II. RELATED WORK The presented state of the art contains practices and models used in successfully applying bio-inspired concepts from self- regulation biological systems to autonomic computing. Membrane computing systems have been successfully ap- plied for solving NP problems in various domains. In [7], a P-system is used to implement a depth-first search for finding the solution to the N-Queens problem. The described P-system is shown to solve the N-queens problem for a 20 X 20 board in approximately 15 seconds. Another approach to search problems using P-systems, Reference [8] presents a local search solution using P-systems, successfully applied to the same N-queens problem, showing that while local search algorithms do not guarantee that a solution can be found, such algorithms use less memory and are well suited for problems with large search space. A different search problem solved using P-systems is presented in [9], where the authors construct a system combining membrane computing and ant colony optimization. 978-1-4673-2952-1/12/$31.00 ©2012 IEEE 281