Many-core Algorithms for Combinatorial Optimization

Combinatorial Optimization is becoming ever more crucial, in these days. From natural sciences to economics, passing through urban centers administration and personnel management, methodologies and algorithms with a strong theoretical background and a consolidated real-word effectiveness is more and more requested, in order to find, quickly, good solutions to complex strategical problems. Resource optimization is, nowadays, a fundamental ground for building the basements of successful projects. From the theoretical point of view, Combinatorial Optimization rests on stable and strong foundations, that allow researchers to face ever more challenging problems. However, from the application point of view, it seems that the rate of theoretical developments cannot cope with that enjoyed by modern hardware technologies, especially with reference to the one of processors industry. In this work we propose new parallel algorithms, designed for exploiting the new parallel architectures available on the market. We found that, exposing the inherent parallelism of some resolution techniques (like Dynamic Programming), the computational benefits are remarkable, lowering the execution times by more than an order of magnitude, and allowing to address instances with dimensions not possible before. We approached four Combinatorial Optimization’s notable problems: Packing Problem, Vehicle Routing Problem, Single Source Shortest Path Problem and a Network Design problem. For each of these problems we propose a collection of effective parallel solution algorithms, either for solving the full problem (Guillotine Cuts and SSSPP) or for enhancing a fundamental part of the solution method (VRP and ND). We endorse our claim by presenting computational results for all problems, either on standard benchmarks from the literature or, when possible, on data from real-world applications, where speed-ups of one order of magnitude are usually attained, not uncommonly scaling up to 40 X factors.

Evolutionary methods for self-organizing cooperation in peer-to-peer networks

The Peer-to-Peer network paradigm is drawing the attention of both final users and researchers for its features. P2P networks shift from the classic client-server approach to a high level of decentralization where there is no central control and all the nodes should be able not only to require services, but to provide them to other peers as well. While on one hand such high level of decentralization might lead to interesting properties like scalability and fault tolerance, on the other hand it implies many new problems to deal with. A key feature of many P2P systems is openness, meaning that everybody is potentially able to join a network with no need for subscription or payment systems. The combination of openness and lack of central control makes it feasible for a user to free-ride, that is to increase its own benefit by using services without allocating resources to satisfy other peers’ requests. One of the main goals when designing a P2P system is therefore to achieve cooperation between users. Given the nature of P2P systems based on simple local interactions of many peers having partial knowledge of the whole system, an interesting way to achieve desired properties on a system scale might consist in obtaining them as emergent properties of the many interactions occurring at local node level. Two methods are typically used to face the problem of cooperation in P2P networks: 1) engineering emergent properties when designing the protocol; 2) study the system as a game and apply Game Theory techniques, especially to find Nash Equilibria in the game and to reach them making the system stable against possible deviant behaviors. In this work we present an evolutionary framework to enforce cooperative behaviour in P2P networks that is alternative to both the methods mentioned above. Our approach is based on an evolutionary algorithm inspired by computational sociology and evolutionary game theory, consisting in having each peer periodically trying to copy another peer which is performing better. The proposed algorithms, called SLAC and SLACER, draw inspiration from tag systems originated in computational sociology, the main idea behind the algorithm consists in having low performance nodes copying high performance ones. The algorithm is run locally by every node and leads to an evolution of the network both from the topology and from the nodes’ strategy point of view. Initial tests with a simple Prisoners’ Dilemma application show how SLAC is able to bring the network to a state of high cooperation independently from the initial network conditions. Interesting results are obtained when studying the effect of cheating nodes on SLAC algorithm. In fact in some cases selfish nodes rationally exploiting the system for their own benefit can actually improve system performance from the cooperation formation point of view. The final step is to apply our results to more realistic scenarios. We put our efforts in studying and improving the BitTorrent protocol. BitTorrent was chosen not only for its popularity but because it has many points in common with SLAC and SLACER algorithms, ranging from the game theoretical inspiration (tit-for-tat-like mechanism) to the swarms topology. We discovered fairness, meant as ratio between uploaded and downloaded data, to be a weakness of the original BitTorrent protocol and we drew inspiration from the knowledge of cooperation formation and maintenance mechanism derived from the development and analysis of SLAC and SLACER, to improve fairness and tackle freeriding and cheating in BitTorrent. We produced an extension of BitTorrent called BitFair that has been evaluated through simulation and has shown the abilities of enforcing fairness and tackling free-riding and cheating nodes.

An evolutionary approach to the design of experiments for combinatorial optimization with an application to enzyme engineering

In a large number of problems the high dimensionality of the search space, the vast number of variables and the economical constrains limit the ability of classical techniques to reach the optimum of a function, known or unknown. In this thesis we investigate the possibility to combine approaches from advanced statistics and optimization algorithms in such a way to better explore the combinatorial search space and to increase the performance of the approaches. To this purpose we propose two methods: (i) Model Based Ant Colony Design and (ii) Naïve Bayes Ant Colony Optimization. We test the performance of the two proposed solutions on a simulation study and we apply the novel techniques on an appplication in the field of Enzyme Engineering and Design.

Parallel Architectures for Many-Core Systems-On-Chip in Deep Sub-Micron Technology

Despite the several issues faced in the past, the evolutionary trend of silicon has kept its constant pace. Today an ever increasing number of cores is integrated onto the same die. Unfortunately, the extraordinary performance achievable by the many-core paradigm is limited by several factors. Memory bandwidth limitation, combined with inefficient synchronization mechanisms, can severely overcome the potential computation capabilities. Moreover, the huge HW/SW design space requires accurate and flexible tools to perform architectural explorations and validation of design choices. In this thesis we focus on the aforementioned aspects: a flexible and accurate Virtual Platform has been developed, targeting a reference many-core architecture. Such tool has been used to perform architectural explorations, focusing on instruction caching architecture and hybrid HW/SW synchronization mechanism. Beside architectural implications, another issue of embedded systems is considered: energy efficiency. Near Threshold Computing is a key research area in the Ultra-Low-Power domain, as it promises a tenfold improvement in energy efficiency compared to super-threshold operation and it mitigates thermal bottlenecks. The physical implications of modern deep sub-micron technology are severely limiting performance and reliability of modern designs. Reliability becomes a major obstacle when operating in NTC, especially memory operation becomes unreliable and can compromise system correctness. In the present work a novel hybrid memory architecture is devised to overcome reliability issues and at the same time improve energy efficiency by means of aggressive voltage scaling when allowed by workload requirements. Variability is another great drawback of near-threshold operation. The greatly increased sensitivity to threshold voltage variations in today a major concern for electronic devices. We introduce a variation-tolerant extension of the baseline many-core architecture. By means of micro-architectural knobs and a lightweight runtime control unit, the baseline architecture becomes dynamically tolerant to variations.