Organizational Unit:
School of Computational Science and Engineering
School of Computational Science and Engineering
Permanent Link
Research Organization Registry ID
Description
Previous Names
Parent Organization
Parent Organization
Organizational Unit
Includes Organization(s)
ArchiveSpace Name Record
Publication Search Results
Now showing
1 - 6 of 6
-
ItemEnergy efficient parallel and distributed simulation(Georgia Institute of Technology, 2019-07-26) Biswas, AradhyaNew challenges and opportunities emerge as computing interacts with our surroundings in unprecedented ways. One of these challenges is the energy consumed by computations and communications. In large cloud-based computing systems, it is a major concern because it forms the largest proportion of the environmental and operational costs of data centers. In mobile systems, it directly impacts battery life. This work focuses on understanding and reducing power and energy consumption of the parallel and distributed execution of discrete event simulations, an area not extensively studied in the past. We first empirically characterize the energy consumption of widely used synchronization algorithms. Then a model and techniques are presented and exercised to create energy profile of a distributed simulation system. These demonstrate that distributed execution and synchronization can incur a significant energy and power overhead. To study and optimize the energy required for distributed execution, a property termed zero-energy synchronization is proposed. A zero-energy synchronization algorithm based on an oracle is presented, and a practical implementation is discussed. A more generic synchronization algorithm termed Low Energy YAWNS (LEY) is also proposed. LEY represents the first attempt to design a synchronization algorithm for energy efficiency and, in principle, can achieve zero-energy synchronization for a large class of distributed simulation applications. To employ the energy efficiency of specialized computing hardware platforms, recurrence relations for simulating G/G/1 queueing networks, directly implementable using library primitives, are proposed. In addition to optimizations and scalability they offer, the use of library primitives ease development and open up avenues for adapting the simulation for custom hardware. Composition of parallel prefix scans further improve the energy efficiency of the proposed recurrences and similar sequences of parallel prefix scans.
-
ItemEnergy efficient data driven distributed traffic simulations(Georgia Institute of Technology, 2018-04-05) Neal, Sabra AlexandriaWith the growing capabilities of the Internet of Things and proliferation of mobile devices interest in the use of real-time data as a means for input to distributed online simulations has increased. Online simulations provide users with the ability to utilize real-time data to make adaptations to the system, e.g., to adjust to unexpected events. One problem that arises when using these systems on mobile devices is that they are dependent upon the device’s stored energy. It is vital to understand how all components of such a system use the stored energy in order to understand how to develop such systems for energy constrained environments. One aspect of this thesis is to examine the role that discrete event driven and cellular automata models have on energy consumption in embedded systems. Discrete event driven simulations are dependent on a future event list for execution. It is important to understand the affect of the data structure for the future event list on energy consumption when running such simulations in embedded systems. This thesis presents a characterization of the relationship between the operations performed on the future event list and energy consumption. This thesis investigates an energy aware approach applicable for systems that are restricted to energy constrained environments.
-
ItemParallel simulation of scale-free networks(Georgia Institute of Technology, 2017-08-01) Nguyen, Thuy Vy ThuyIt has been observed that many networks arising in practice have skewed node degree distributions. Scale-free networks are one well-known class of such networks. Achieving efficient parallel simulation of scale-free networks is challenging because large-degree nodes can create bottlenecks that limit performance. To help address this problem, we describe an approach called link partitioning where each network link is mapped to a logical process in contrast to the conventional approach of mapping each node to a logical process.
-
ItemVirtual time-aware virtual machine systems(Georgia Institute of Technology, 2014-07-01) Yoginath, Srikanth B.Discrete dynamic system models that track, maintain, utilize, and evolve virtual time are referred to as virtual time systems (VTS). The realization of VTS using virtual machine (VM) technology offers several benefits including fidelity, scalability, interoperability, fault tolerance and load balancing. The usage of VTS with VMs appears in two ways: (a) VMs within VTS, and (b) VTS over VMs. The former is prevalent in high-fidelity cyber infrastructure simulations and cyber-physical system simulations, wherein VMs form a crucial component of VTS. The latter appears in the popular Cloud computing services, where VMs are offered as computing commodities and the VTS utilizes VMs as parallel execution platforms. Prior to our work presented here, the simulation community using VM within VTS (specifically, cyber infrastructure simulations) had little awareness of the existence of a fundamental virtual time-ordering problem. The correctness problem was largely unnoticed and unaddressed because of the unrecognized effects of fair-share multiplexing of VMs to realize virtual time evolution of VMs within VTS. The dissertation research reported here demonstrated the latent incorrectness of existing methods, defined key correctness benchmarks, quantitatively measured the incorrectness, proposed and implemented novel algorithms to overcome incorrectness, and optimized the solutions to execute without a performance penalty. In fact our novel, correctness-enforcing design yields better runtime performance than the traditional (incorrect) methods. Similarly, the VTS execution over VM platforms such as Cloud computing services incurs large performance degradation, which was not known until our research uncovered the fundamental mismatch between the scheduling needs of VTS execution and those of traditional parallel workloads. Consequently, we designed a novel VTS-aware hypervisor scheduler and showed significant performance gains in VTS execution over VM platforms. Prior to our work, the performance concern of VTS over VM was largely unaddressed due to the absence of an understanding of execution policy mismatch between VMs and VTS applications. VTS follows virtual-time order execution whereas the conventional VM execution follows fair-share policy. Our research quantitatively uncovered the exact cause of poor performance of VTS in VM platforms. Moreover, we proposed and implemented a novel virtual time-aware execution methodology that relieves the degradation and provides over an order of magnitude faster execution than the traditional virtual time-unaware execution.
-
ItemAd hoc distributed simulation: a method for embedded online simulations(Georgia Institute of Technology, 2013-07-02) Huang, Ya-LinThe continual growth of computing power in small devices has motivated the development of novel approaches to optimizing operational systems efficiently and effectively. These optimization problems are often so complex that solving them analytically may be difficult, if not prohibited. One method for solving such problems is to use online simulation. However, challenges in using online simulation include the issues of responsiveness (e.g., because of communication delays), scalability, and failure resistance. To tackle these issues, this study proposes embedding online simulations into a network of sensors that monitors the system under investigation. This thesis explores an approach termed “ad hoc distributed simulation,” which is based on embedding online simulations into a sensor network and adding communication and synchronization among simulators to model operational systems. This approach offers several potential advantages over existing approaches: (1) it can provide rapid response to system dynamics as well as efficiency since data exchange is local to the sensor network, (2) it can achieve better scalability to incorporate more sensors, and (3) it can provide better robustness to failures because portions of the system are still under local control. This research addresses several statistical issues in this ad hoc approach: (1) rapid and effective estimation of the input processes at model boundaries, (2) estimation of system-wide performance measures from individual simulator outputs, and (3) correction mechanisms responding to unexpected events or inaccuracies within the model. This thesis examines ad hoc distributed simulation analytically and experimentally, mainly focusing on the accuracy of predicting the performance of open queueing networks. First, the analytical part formalizes the ad hoc approach and evaluates its accuracy at modeling certain class of open queueing networks with regard to the steady-state system performance measures. This work concerning steady-state metrics is extended to a broader class of networks by an empirical study, which presents evidence to show that the ad hoc approach can generate predictions comparable to those from sequential simulations. Furthermore, a “buffered-area” mechanism is proposed to substantially reduce prediction bias with a moderate increase in execution time. In addition to those steady-state studies, another empirical study targets the prediction accuracy of the ad hoc approach at open queueing networks with short-term system-state transients. This study demonstrates that, with slight modification to the prior design of the ad hoc queueing simulation method for those steady-state studies, system dynamics can be well modeled. The results, again, support the conclusion that the ad hoc approach is competitive to the sequential simulation method in terms of prediction accuracy.
-
ItemInterest management scheme and prediction model in intelligent transportation systems(Georgia Institute of Technology, 2012-10-12) Li, YingThis thesis focuses on two important problems related to DDDAS: interest management (data distribution) and prediction models. In order to reduce communication overhead, we propose a new interest management mechanism for mobile peer-to-peer systems. This approach involves dividing the entire space into cells and using an efficient sorting algorithm to sort the regions in each cell. A mobile landmarking scheme is introduced to implement this sort-based scheme in mobile peer-to-peer systems. The design does not require a centralized server, but rather, every peer can become a mobile landmark node to take a server-like role to sort and match the regions. Experimental results show that the scheme has better computational efficiency for both static and dynamic matching. In order to improve communication efficiency, we present a travel time prediction model based on boosting, an important machine learning technique, and combine boosting and neural network models to increase prediction accuracy. We also explore the relationship between the accuracy of travel time prediction and the frequency of traffic data collection with the long term goal of minimizing bandwidth consumption. Several different sets of experiments are used to evaluate the effectiveness of this model. The results show that the boosting neural network model outperforms other predictors.