Meta-models, environment and layers: agent-oriented engineering of complex systems

Traditional software engineering approaches and metaphors fall short when applied to areas of growing relevance such as electronic commerce, enterprise resource planning, and mobile computing: such areas, in fact, generally call for open architectures that may evolve dynamically over time so as to accommodate new components and meet new requirements. This is probably one of the main reasons that the agent metaphor and the agent-oriented paradigm are gaining momentum in these areas. This thesis deals with the engineering of complex software systems in terms of the agent paradigm. This paradigm is based on the notions of agent and systems of interacting agents as fundamental abstractions for designing, developing and managing at runtime typically distributed software systems. However, today the engineer often works with technologies that do not support the abstractions used in the design of the systems. For this reason the research on methodologies becomes the basic point in the scientific activity. Currently most agent-oriented methodologies are supported by small teams of academic researchers, and as a result, most of them are in an early stage and still in the first context of mostly \academic" approaches for agent-oriented systems development. Moreover, such methodologies are not well documented and very often defined and presented only by focusing on specific aspects of the methodology. The role played by meta- models becomes fundamental for comparing and evaluating the methodologies. In fact a meta-model specifies the concepts, rules and relationships used to define methodologies. Although it is possible to describe a methodology without an explicit meta-model, formalising the underpinning ideas of the methodology in question is valuable when checking its consistency or planning extensions or modifications. A good meta-model must address all the different aspects of a methodology, i.e. the process to be followed, the work products to be generated and those responsible for making all this happen. In turn, specifying the work products that must be developed implies dening the basic modelling building blocks from which they are built. As a building block, the agent abstraction alone is not enough to fully model all the aspects related to multi-agent systems in a natural way. In particular, different perspectives exist on the role that environment plays within agent systems: however, it is clear at least that all non-agent elements of a multi-agent system are typically considered to be part of the multi-agent system environment. The key role of environment as a first-class abstraction in the engineering of multi-agent system is today generally acknowledged in the multi-agent system community, so environment should be explicitly accounted for in the engineering of multi-agent system, working as a new design dimension for agent-oriented methodologies. At least two main ingredients shape the environment: environment abstractions - entities of the environment encapsulating some functions -, and topology abstractions - entities of environment that represent the (either logical or physical) spatial structure. In addition, the engineering of non-trivial multi-agent systems requires principles and mechanisms for supporting the management of the system representation complexity. These principles lead to the adoption of a multi-layered description, which could be used by designers to provide different levels of abstraction over multi-agent systems. The research in these fields has lead to the formulation of a new version of the SODA methodology where environment abstractions and layering principles are exploited for en- gineering multi-agent systems.

Computing the structural and vibrational properties of polymorphic organic molecular crystals through van der Waals corrected density functional theory and the electronic properties of organic thin films through microelectrostatic calculations.

The present Thesis reports on the various research projects to which I have contributed during my PhD period, working with several research groups, and whose results have been communicated in a number of scientific publications. The main focus of my research activity was to learn, test, exploit and extend the recently developed vdW-DFT (van der Waals corrected Density Functional Theory) methods for computing the structural, vibrational and electronic properties of ordered molecular crystals from first principles. A secondary, and more recent, research activity has been the analysis with microelectrostatic methods of Molecular Dynamics (MD) simulations of disordered molecular systems. While only very unreliable methods based on empirical models were practically usable until a few years ago, accurate calculations of the crystal energy are now possible, thanks to very fast modern computers and to the excellent performance of the best vdW-DFT methods. Accurate energies are particularly important for describing organic molecular solids, since they often exhibit several alternative crystal structures (polymorphs), with very different packing arrangements but very small energy differences. Standard DFT methods do not describe the long-range electron correlations which give rise to the vdW interactions. Although weak, these interactions are extremely sensitive to the packing arrangement, and neglecting them used to be a problem. The calculations of reliable crystal structures and vibrational frequencies has been made possible only recently, thanks to development of some good representations of the vdW contribution to the energy (known as “vdW corrections”).

Monitoring and Prediction of Thermal Emergencies in High Performance Computing Systems

Modern scientific discoveries are driven by an unsatisfiable demand for computational resources. High-Performance Computing (HPC) systems are an aggregation of computing power to deliver considerably higher performance than one typical desktop computer can provide, to solve large problems in science, engineering, or business. An HPC room in the datacenter is a complex controlled environment that hosts thousands of computing nodes that consume electrical power in the range of megawatts, which gets completely transformed into heat. Although a datacenter contains sophisticated cooling systems, our studies indicate quantitative evidence of thermal bottlenecks in real-life production workload, showing the presence of significant spatial and temporal thermal and power heterogeneity. Therefore minor thermal issues/anomalies can potentially start a chain of events that leads to an unbalance between the amount of heat generated by the computing nodes and the heat removed by the cooling system originating thermal hazards. Although thermal anomalies are rare events, anomaly detection/prediction in time is vital to avoid IT and facility equipment damage and outage of the datacenter, with severe societal and business losses. For this reason, automated approaches to detect thermal anomalies in datacenters have considerable potential. This thesis analyzed and characterized the power and thermal characteristics of a Tier0 datacenter (CINECA) during production and under abnormal thermal conditions. Then, a Deep Learning (DL)-powered thermal hazard prediction framework is proposed. The proposed models are validated against real thermal hazard events reported for the studied HPC cluster while in production. This thesis is the first empirical study of thermal anomaly detection and prediction techniques of a real large-scale HPC system to the best of my knowledge. For this thesis, I used a large-scale dataset, monitoring data of tens of thousands of sensors for around 24 months with a data collection rate of around 20 seconds.

Flexible Computing Systems For AI Acceleration At The Extreme Edge Of The IoT

Embedding intelligence in extreme edge devices allows distilling raw data acquired from sensors into actionable information, directly on IoT end-nodes. This computing paradigm, in which end-nodes no longer depend entirely on the Cloud, offers undeniable benefits, driving a large research area (TinyML) to deploy leading Machine Learning (ML) algorithms on micro-controller class of devices. To fit the limited memory storage capability of these tiny platforms, full-precision Deep Neural Networks (DNNs) are compressed by representing their data down to byte and sub-byte formats, in the integer domain. However, the current generation of micro-controller systems can barely cope with the computing requirements of QNNs. This thesis tackles the challenge from many perspectives, presenting solutions both at software and hardware levels, exploiting parallelism, heterogeneity and software programmability to guarantee high flexibility and high energy-performance proportionality. The first contribution, PULP-NN, is an optimized software computing library for QNN inference on parallel ultra-low-power (PULP) clusters of RISC-V processors, showing one order of magnitude improvements in performance and energy efficiency, compared to current State-of-the-Art (SoA) STM32 micro-controller systems (MCUs) based on ARM Cortex-M cores. The second contribution is XpulpNN, a set of RISC-V domain specific instruction set architecture (ISA) extensions to deal with sub-byte integer arithmetic computation. The solution, including the ISA extensions and the micro-architecture to support them, achieves energy efficiency comparable with dedicated DNN accelerators and surpasses the efficiency of SoA ARM Cortex-M based MCUs, such as the low-end STM32M4 and the high-end STM32H7 devices, by up to three orders of magnitude. To overcome the Von Neumann bottleneck while guaranteeing the highest flexibility, the final contribution integrates an Analog In-Memory Computing accelerator into the PULP cluster, creating a fully programmable heterogeneous fabric that demonstrates end-to-end inference capabilities of SoA MobileNetV2 models, showing two orders of magnitude performance improvements over current SoA analog/digital solutions.

PCM-based in-memory computing: architectures, circuits and applications

Analog In-memory Computing (AIMC) has been proposed in the context of Beyond Von Neumann architectures as a valid strategy to reduce internal data transfers energy consumption and latency, and to improve compute efficiency. The aim of AIMC is to perform computations within the memory unit, typically leveraging the physical features of memory devices. Among resistive Non-volatile Memories (NVMs), Phase-change Memory (PCM) has become a promising technology due to its intrinsic capability to store multilevel data. Hence, PCM technology is currently investigated to enhance the possibilities and the applications of AIMC. This thesis aims at exploring the potential of new PCM-based architectures as in-memory computational accelerators. In a first step, a preliminar experimental characterization of PCM devices has been carried out in an AIMC perspective. PCM cells non-idealities, such as time-drift, noise, and non-linearity have been studied to develop a dedicated multilevel programming algorithm. Measurement-based simulations have been then employed to evaluate the feasibility of PCM-based operations in the fields of Deep Neural Networks (DNNs) and Structural Health Monitoring (SHM). Moreover, a first testchip has been designed and tested to evaluate the hardware implementation of Multiply-and-Accumulate (MAC) operations employing PCM cells. This prototype experimentally demonstrates the possibility to reach a 95% MAC accuracy with a circuit-level compensation of cells time drift and non-linearity. Finally, empirical circuit behavior models have been included in simulations to assess the use of this technology in specific DNN applications, and to enhance the potentiality of this innovative computation approach.