Vivek Sarkar, Zoran Budimlić, Milind Kulkani · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 2016
This report summarizes runtime system challenges for exascale computing, that follow from the fundamental challenges for exascale systems that have been well studied in past reports, e.g., [6, 33, 34, 32, 24]. Some of the key exascale challenges that pertain to runtime systems include parallelism, energy efficiency, memory hierarchies, data movement, heterogeneous processors and memories, resilience, performance variability, dynamic resource allocation, performance portability, and interoperability with legacy code. In addition to summarizing these challenges, the report also outlines different approaches to addressing these significant challenges that have been pursued by research projects in the DOE-sponsored X-Stack and OS/R programs. Since there is often confusion as to what exactly the term “runtime system” refers to in the software stack, we include a section on taxonomy to clarify the terminology used by participants in these research projects. In addition, we include a section on deployment opportunities for vendors and government labs to build on the research results from these projects. Finally, this report is also intended to provide a framework for discussing future research and development investments for exascale runtime systems, and for clarifying the role of runtime systems in exascale software.
Cited by 1Open access
Lethin, Richard, Sarkar, Vivek, Budimlic, Zoran et al. · 2016
Cited by 0Open access
Jason Gregory · 2026
Cited by 0
2010
(FASTOS). This announcement is focused on research and development of operating and runtime systems which enable the effective management and use of extreme-scale systems (petascale and beyond) for scientific computation. The overall goal of this announcement is to stimulate research and development related to operating and runtime systems for petascale systems the in 2010 to 2015 timeframe. It is likely that these systems will include a combination of commodity and custom components, with different systems reflecting different degrees of customization. Operating and runtime systems research must be driven from the needs of current and future applications, and the primary focus is on supporting the needs of existing and anticipated SC and other DOE applications. An ultimate goal would be the development of a unified operating and runtime system that could fully support and exploit petascale and beyond systems and autonomously adapt to meet specific application needs for performance, functionality, security, and fault tolerance. The activities supported by this notice may be a combination of basic research, development, prototyping, and testing. Partnerships among universities, National Laboratories, and industry are encouraged. PREAPPLICATION DUE DATE: April 6, 2007, 4:30 pm, Eastern Time Potential applicants are required to submit a two-page preapplication by email to
Cited by 0
2007
Cited by 0
Gruhn, Volker, Schäfer, Clemens · Publikation Server (Leipzig University)
Cited by 0
2014
Cited by 0
Ganguly, Ritam · Michigan State University Libraries · 2024
Cited by 0Open access
Volker Gruhn, Sch auml fer Clemens · Frontiers in artificial intelligence and applications · 2008
The emerging behavior of a mobile system–where hardware devices of different types interoperate–is determined by its software architecture (structure, dynamics, deployment), the underlying communication networks (topology, properties like bandwidth etc.) and interactions undertaken by the users of the system. In order to assess whether a mobile system fulfills its non-functional requirements like response times or availability already at design time, the emergent behavior of such a system can be simulated by using an architectural model of the system and applying an simulation approach where a network model and a user interaction model are used for providing the contextual information.
Cited by 0
David Griffiths, Tom Franklin, Paul Sharples et al. · DSpace (Open University in the Netherlands) · 2008
This Annex provides detailed information on work carried out on runtime systems during the first 18 months of the TENCompetence project
Cited by 0
Alexandre David, Kim G. Larsen, Axel Legay et al. · Lecture notes in computer science · 2012
Cited by 27Open access
Antoine Bernabeu · SPIRE - Sciences Po Institutional REpository · 2023
REMERCIEMENTSLes travaux prsents dans ce mmoire ont t effectus au sein du laboratoire des sciences du numrique de Nantes (LS2N).Je remercie le personnel du laboratoire pour m'avoir accueilli.Je remercie particulirement mes encadrants de thse,
Cited by 0Open access
Simon Fau · theses.fr (ABES) · 2016
Notre approche dans cette thèse était d'identifier où le chiffrement complètement homomorphe (FHE) pouvait être utilisé pour le domaine des sciences informatiques et de construire une plate-forme expérimentale qui nous permette de tester des algorithmes de traitement de l'information manipulant des données chiffrées. La première partie de cette thèse est consacrée à l'état de l'art. Nous présentons d'abord les systèmes de chiffrement homomorphes conçus avant 2008, puis nous présentons ceux adressant la problématique du chiffrement complètement homomorphe. Nous décrivons plusieurs méthodes de chiffrement d'intérêt pour cette thèse et discutons de leurs implémentations FHE. Enfin, nous présentons des circuits de Yao car ils peuvent résoudre des problèmes similaires que le FHE et nous parlons brièvement du chiffrement fonctionnel (FE). La deuxième partie de cette thèse présente nos contributions. Nous commençons par expliquer comment le FHE peut être utile dans divers scénarios et décrivons plusieurs cas d'utilisation pratique identifiés au cours de la thèse. Ensuite, nous décrivons notre approche pour effectuer des calculs sur des données chiffrées à l'aide du FHE et expliquons comment nous avons pu développer une plate-forme pour l'exécution dans le domaine chiffré d'une large gamme d'algorithmes en s'appuyant seulement sur l'addition et la multiplication homomorphes. Nous détaillons ensuite notre solution pour effectuer des requêtes privées sur une base de données chiffrées en utilisant le chiffrement homomorphe. Dans un dernier chapitre, nous présentons nos résultats expérimentaux.
Cited by 0Open access
Bouge, Luc, Mehaut, Jean-Francois, Namyst, Raymond et al. · OpenGrey (Institut de l'Information Scientifique et Technique) · 1999
Cited by 0
Vojtech Forejt, Marta Kwiatkowska, David Parker et al. · Oxford University Research Archive (ORA) (University of Oxford) · 2012
Probabilistic verification techniques have been proposed for runtime analysis of adaptive software systems, with the verification results being used to steer the system so that it satisfies certain Quality-of-Service requirements. Since systems evolve over time, and verification results are required promptly, efficiency is an essential issue. To address this, we present incremental verification techniques, which exploit the results of previous analyses. We target systems modelled as Markov decision processes, developing incremental methods for constructing models from high-level system descriptions and for numerical solution using policy iteration based on strongly connected components. A prototype implementation, based on the PRISM model checker, demonstrates performance improvements on a range of case studies.
Cited by 0Open access
Babaee Cheshmeahmadrezaee, Reza · UWSpace (University of Waterloo) · 2019
Runtime Verification (RV) is the formal analysis of the execution of a system against some \nproperties at runtime. RV is particularly useful for stochastic systems that have a non-zero \nprobability of failure at runtime. The standard RV assumes constructing a monitor that \nchecks only the currently observed execution of the system against the given properties. \nThis dissertation proposes a framework for predictive RV, where the monitor instead \nchecks the current execution with its finite extensions against some property. The extensions are generated using a prediction model, that is built based on execution samples \nrandomly generated from the system. The thesis statement is that predictive RV for \nstochastic systems is feasible, effective, and useful. \nThe feasibility is demonstrated by providing a framework, called Prevent, that builds a \npredictive monitor by using trained prediction models to finitely extend an execution path, \nand computing the probabilities of the extensions that satisfy or violate the given property. \nThe prediction model is trained using statistical learning techniques from independent and \nidentically distributed samples of system executions. The prediction is the result of a \nquantitative bounded reachability analysis on the product of the prediction model and the \nautomaton specifying the property. The analysis results are computed offline and stored in \na lookup table. At runtime the monitor obtains the state of the system on the prediction \nmodel based on the observed execution, directly or by approximation, and uses the lookup \ntable to retrieve the computed probability that the system at the current state will satisfy \nor violate the given property within some finite number of steps. \nThe effectiveness of Prevent is shown by applying abstraction when constructing the \nprediction model. The abstraction is on the observation space based on extracting the \nsymmetry relation between symbols that have similar probabilities to satisfy a property. \nThe abstraction may introduce nondeterminism in the final model, which is handled by \nusing a hidden state variable when building the prediction model. We also demonstrate \nthat, under the convergence conditions of the learning algorithms, the prediction results \nfrom the abstract models are the same as the concrete models. \nFinally, the usefulness of Prevent is indicated in real-world applications by showing \nhow it can be applied for predicting rare properties, properties with very low but non-zero \nprobability of satisfaction. More specifically, we adjust the training algorithm that uses \nthe samples generated by importance sampling to generate the prediction models for rare \nproperties without increasing the number of samples and without having a negative impact \non the prediction accuracy.
Cited by 0Open access
Baskaran Krishnamoorthy · 2005
Reconfigurable computing combines the benefits of both software and reconfigurable hardware implementations. The use of FPGA in reconfigurable computing is becoming increasingly prevalent due to its ever-increasing density and rapid turn around time.
Cited by 1Open access
ENTRUST Horizon Europe · Zenodo (CERN European Organization for Nuclear Research) · 2024
EuroSys '24: Proceedings of the Nineteenth European Conference on Computer Systems Pages 1175 - 1191 https://doi.org/10.1145/3627703.3650070 ABSTRACT Batteryless energy harvesting devices compute intermittently due to power failures that frequently interrupt the computational activity and lead to charging delays. To ensure functional correctness in intermittent computing, applications must exhibit several unique properties, such as guarantees for computational progress despite power failures and prevention of stale operations caused by charging delays. We observe that current software support for intermittent computing allows for checking only a fixed set of properties and leads to tightly coupled application and property-checking, thus hampering modularity, scalability, and maintainability. In this paper, we present ARTEMIS, the first framework designed to facilitate flexible property checking of intermittent programs at runtime. ARTEMIS is developed based on techniques from the area of runtime monitoring, offers a specification language for specifying an open set of properties, and provides automatic generation of monitors responsible for checking the properties. Our evaluation showed that ARTEMIS achieves comparable efficiency to state-of-the-art solutions while significantly preventing failure scenarios through its monitoring capabilities. AUTHORS Eren Yildiz, Khakim Akhunov, Lorenzo Antonio Riva, Arda Goknil, Ivan Kurtev, Kasim Sinan Yildirim
Cited by 0Open access
Sanjay Chatterjee · 2013
Abstract : Future extreme-scale systems are expected to contain homogeneous and heterogeneous many-core processors, with O(103) cores per node and O(106) nodes overall. Effective combination of inter-node and intra-node parallelism is recognized to be a major software challenge for such systems. Further, applications will have to deal with constrained energy budgets as well as frequent faults and failures. To aid programmers manage these complexities and enhance programmability, much of recent research has focused on designing state-of-art software runtime systems. Such runtime systems are expected to be a critical component of the software ecosystem for the management of parallelism, locality, load balancing, energy and resilience on extreme-scale systems. In this dissertation, we address three key challenges faced by a runtime system using a dynamic task parallel framework for extreme-scale computing.
Cited by 1
Chatterjee, Sanjay · Rice Digital Scholarship Archive (Rice University) · 2014
Future extreme-scale systems are expected to contain homogeneous and heterogeneous\nmany-core processors, with up to $O(10^3)$ cores per node and $O(10^6)$ nodes overall.\nEffective combination of inter-node and intra-node parallelism is recognized to\nbe a major software challenge for such systems.\nDynamic task parallelism, which forms the basis of our research, has proven to\nbe an effective execution model enabled by efficient runtime systems for scalable\nperformance on shared-memory programs.\nThis work builds on the Habanero-C task parallel language.\nWe focus on runtime execution models for scalable locality aware task scheduling\nand synchronization at both intra-node and inter-node levels.\n\nA scalable runtime communication system is a critical requirement for extreme-scale computing.\nWe present our runtime communication system, called HC-COMM, which can integrate\nany popular communication system, such as MPI, with dynamic asynchronous tasks.\nThe runtime uses dedicated computation and communication cores for scalable performance.\nWe introduce the HC-MPI execution model, where all MPI calls are treated as\nasynchronous tasks, thereby enabling unified handling of messages and tasking constructs.\nHC-MPI provides easy portability of existing MPI applications.\nWe also introduce the HAPGNS programming model that uses the HC-COMM runtime\nto extend shared-memory data flow programming onto distributed-memory.\nIn this model, the user does not require any knowledge of MPI.\nWe believe locality of computation and data will be critical for higher performance\nand lower energy resulting from data reuse on faster memories and redundant\ncommunication avoidance.\nOur locality optimization framework can enable performance experts to optimize\nexisting applications via tuning operations.\nThese optimizations aim to benefit from spatial and temporal task locality using\nruntime co-scheduling of tasks on a hierarchical place tree construct which\nmodels the system memory hierarchy.\nIn the HAPGNS model, user provided data distribution functions act as data\nlocality directives at inter-node level.\nFinally, we present efficient algorithms to enable the Phaser synchronization\nmodel on long running tasks.\nWe have extended the Phaser construct to support unified collective operations\nat both inter-node and intra-node levels.\nOur experimental results on current large scale systems serve to validate the\ndesign and implementation of our runtime system for extreme-scale computing.
Cited by 0