Dynamic Partitioning Scheme Research Articles

Multi-Agent Deep Reinforcement Learning Based Dynamic Task Offloading in a Device-to-Device Mobile-Edge Computing Network to Minimize Average Task Delay with Deadline Constraints.

Device-to-device (D2D) is a pivotal technology in the next generation of communication, allowing for direct task offloading between mobile devices (MDs) to improve the efficient utilization of idle resources. This paper proposes a novel algorithm for dynamic task offloading between the active MDs and the idle MDs in a D2D-MEC (mobile edge computing) system by deploying multi-agent deep reinforcement learning (DRL) to minimize the long-term average delay of delay-sensitive tasks under deadline constraints. Our core innovation is a dynamic partitioning scheme for idle and active devices in the D2D-MEC system, accounting for stochastic task arrivals and multi-time-slot task execution, which has been insufficiently explored in the existing literature. We adopt a queue-based system to formulate a dynamic task offloading optimization problem. To address the challenges of large action space and the coupling of actions across time slots, we model the problem as a Markov decision process (MDP) and perform multi-agent DRL through multi-agent proximal policy optimization (MAPPO). We employ a centralized training with decentralized execution (CTDE) framework to enable each MD to make offloading decisions solely based on its local system state. Extensive simulations demonstrate the efficiency and fast convergence of our algorithm. In comparison to the existing sub-optimal results deploying single-agent DRL, our algorithm reduces the average task completion delay by 11.0% and the ratio of dropped tasks by 17.0%. Our proposed algorithm is particularly pertinent to sensor networks, where mobile devices equipped with sensors generate a substantial volume of data that requires timely processing to ensure quality of experience (QoE) and meet the service-level agreements (SLAs) of delay-sensitive applications.

Heterogeneous cognitive learning particle swarm optimization for large-scale optimization problems

Large-scale optimization problems (LSOPs) become increasingly ubiquitous but complicated in real-world scenarios. Confronted with such sophisticated optimization problems, most existing optimizers dramatically lose their effectiveness. To tackle this type of problems effectively, we propose a heterogeneous cognitive learning particle swarm optimizer (HCLPSO). Unlike most existing particle swarm optimizers (PSOs), HCLPSO partitions particles in the current swarm into two categories, namely superior particles (SP) and inferior particles (IP), based on their fitness, and then treats the two categories of particles differently. For inferior particles, this paper devises a random elite cognitive learning (RECL) strategy to update each one with a random superior particle chosen from SP. For superior particles, this paper designs a stochastic dominant cognitive learning (SDCL) strategy to evolve each one by randomly selecting one guiding exemplar from SP and then updating it only when the selected exemplar is better. With the collaboration between these two learning mechanisms, HCLPSO expectedly evolves particles effectively to explore the search space and exploit the found optimal zones appropriately to find optimal solutions to LSOPs. Furthermore, to help HCLPSO traverse the vast search space with promising compromise between intensification and diversification, this paper devises a dynamic swarm partition scheme to dynamically separate particles into the two categories. With this dynamic strategy, HCLPSO gradually switches from exploring the search space to exploiting the found optimal zones intensively. Experiments are executed on the publicly acknowledged CEC2010 and CEC2013 LSOP benchmark suites to compare HCLPSO with several state-of-the-art approaches. Experimental results reveal that HCLPSO is effective to tackle LSOPs, and attains considerably competitive or even far better optimization performance than the compared state-of-the-art large-scale methods. Furthermore, the effectiveness of each component in HCLPSO and the good scalability of HCLPSO are also experimentally verified.

CASHT: Contention Analysis in Shared Hierarchies with Thefts

Cache management policies should consider workloads’ contention behavior when managing a shared cache. Prior art makes estimates about shared cache behavior by adding extra logic or time to isolate per workload cache statistics. These approaches provide per-workload analysis but do not provide a holistic understanding of the utilization and effectiveness of caches under the ever-growing contention that comes standard with scaling cores. We present Contention Analysis in Shared Hierarchies using Thefts, or CASHT, 1 a framework for capturing cache contention information both offline and online. CASHT takes advantage of cache statistics made richer by observing a consequence of cache contention: inter-core evictions, or what we call THEFTS. We use thefts to complement more familiar cache statistics to train a learning model based on Gradient-boosting Trees (GBT) to predict the best ways to partition the last-level cache. GBT achieves 90+% accuracy with trained models as small as 100 B and at least 95% accuracy at 1 kB model size when predicting the best way to partition two workloads. CASHT employs a novel run-time framework for collecting thefts-based metrics despite partition intervention, and enables per-access sampling rather than set sampling that could add overhead but may not capture true workload behavior. Coupling CASHT and GBT for use as a dynamic policy results in a very lightweight and dynamic partitioning scheme that performs within a margin of error of Utility-based Cache Partitioning at a 1/8 the overhead.

Elastic Collision Based Dynamic Partitioning Scheme for Hybrid Simulations.

In hybrid simulations, such as the QM/MM approach, the system is partitioned into regions that are treated at different levels of theory. The key question then becomes how to evaluate the interactions between particles on opposite sides of the boundary. One approach is to place the boundary in such a way that particles near the boundary on both sides are of the same type, thus simplifying the evaluation of the interactions. If mobile particles are present, such as solvent molecules, and particles are allowed to cross the boundary, the conservation of energy and atomic forces is problematic unless the computational effort is increased significantly. By preventing particles from crossing the boundary but allowing the boundary to be flexible, an accurate estimate of average thermodynamic properties is obtained in principle as illustrated by the flexible inner region ensemble separator (FIRES) method [C. Rowley and B. Roux, J. Chem. Theory Comput. 2012, 8, 3526]. In FIRES, a harmonic restraint is applied to particles near the boundary. Therefore, it can occur that particle cross the boundary to some extent resulting in anomalies in the particle density. Here, a constraint approach is presented where particles instantaneously scatter from the boundary. This scattering-adapted FIRES (SAFIRES) implementation makes use of a variable-time-step propagation algorithm where the time step is scaled automatically to identify the moment a collision should occur. If the length of the time step is kept constant, this propagator reduces to a regular Langevin dynamics algorithm, and to the velocity Verlet algorithm for conservative dynamics if the friction coefficient is set to zero. Correct average ensemble statistics are obtained as demonstrated in simulations where, for testing purposes, the particles in the two regions are treated at the same level of theory, namely, a homogeneous Lennard-Jones (LJ) liquid and liquid water based on the TIP4P potential function. In order to illustrate this approach in solid-liquid interface simulations, a LJ liquid in contact with the surface of a crystal is also simulated. The simulations using SAFIRES are shown to reproduce the unconstrained reference simulations without significant deviations in the particle density and the dynamics are shown to conserve energy when coupling to the heat bath is turned off.

SwMR: A Framework for Accelerating MapReduce Applications on Sunway Taihulight

MapReduce has already been an indispensable framework for the programmers to develop big data applications at scale without concerning the underlying complex system details. However, such an important framework is missing on the Sunway many-core processor that powers the world-leading supercomputer Sunway Taihulight. This paper fills the gap by implementing an efficient MapReduce framework on Sunway many-core processor which takes full advantage of the architecture features such as local device memory and register communication. Specifically, we propose three different schemes that adopt different methods to partition the computation of map and reduce across the MPE and CPEs of Sunway processor. Especially in CPE-accelerated Dynamic Partitioning Scheme (CDPS), the processing role of the CPEs within each CPE pair can be transformed dynamically between map and reduce, which is effective to improve the load balance during runtime. In addition, we adopt the local device memory as well as register communication on each CPE to improve the efficiency of data access and communication. The experiment results demonstrate our MapReduce framework based on CDPS achieves 36.9× performance speedup on average across representative benchmarks.

Dynamic Preamble-Resource Partitioning for Critical MTC in Massive MIMO Systems

Preamble resources are scarce and precious in random access (RA), which need to be efficiently utilized to support critical machine-type communication (MTC) with stringent access requirements. In this article, we study a grant-free RA scenario for critical MTC in the context of massive multiple-input–multiple-output (MIMO). To enhance the access reliability of delay-sensitive devices within a predefined latency budget, two dynamic preamble-resource partitioning (DPP) schemes are proposed. Particularly, by leveraging massive MIMO, we first analytically investigate the feasibility of DPP in the considered RA scenario and demonstrate its performance superiority to the conventional scheme. Based on the analytical results, we then propose a greedy DPP scheme that performs locally optimal preamble-resource partitioning in each RA slot. To find the globally optimal DPP solution, we further propose a reinforcement learning (RL)-based scheme by modeling the considered RA scenario as a Markov decision process. Simulation results show the practicality and effectiveness of the proposed DPP schemes. In particular, to achieve a target access failure rate of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1\times 10^{-2}$ </tex-math></inline-formula> , the proposed RL-based scheme is able to enhance the RA traffic load by 42% and improve the preamble resource utilization by over 25% compared to the conventional baseline scheme.

Dynamic Graph Partitioning Scheme for Supporting Load Balancing in Distributed Graph Environments

As dynamic graph data have been actively used, incremental graph partition schemes have been studied to efficiently store and manage large graphs. In this paper, we propose a vertex-cut based novel incremental graph partitioning scheme that supports load balancing in a distributed environment. The proposed scheme chooses the load of each node that considers its storage utilization and throughput as the partitioning criterion. The proposed scheme defines hot data that means a particular vertex frequently searched among graphs requested by queries. We manage and utilize hot data for graph partitioning. Finally, we perform vertex-cut based dynamic graph partitioning by using a vertex replication index, the load each node, and hot data to distribute the load evenly in a distributed environment. In order to verify the superiority of the proposed partitioning scheme, we compare it with the existing partitioning schemes through a variety of performance evaluations.

OPTIMUS: A Security-Centric Dynamic Hardware Partitioning Scheme for Processors that Prevent Microarchitecture State Attacks

Hardware virtualization allows multiple security-critical and ordinary (insecure) processes to co-execute on a processor. These processes temporally share hardware resources and endure numerous security threats on the microarchitecture state. State-of-the-art secure processor architectures, such as MI6 and IRONHIDE enable capabilities to execute security-critical processes in hardware isolated enclaves utilizing the strong isolation security primitive. The MI6 processor purges small state resources on each enclave entry/exit and statically partitions the last-level cache and DRAM regions to ensure strong isolation. IRONHIDE takes a spatial approach and creates two isolated clusters of cores in a multicore processor to ensure strong isolation for processes executing in the enclave cluster. Both architectures observe performance degradation due to static partitioning of shared hardware resources. OPTIMUS proposes a security-centric dynamic hardware resource partitioning scheme that operates entirely at runtime and ensures strong isolation. It enables deterministic resource allocations at the application level granularity, and limits the number of hardware reconfigurations to ensure bounded information leakage via the timing and termination channels. The dynamic hardware resource partitioning capability of OPTIMUS is shown to co-optimize performance and security for the MI6 and IRONHIDE architectures.

ECI-Cache

In recent years, high interest in using Virtual Machines (VMs) in data centers and Cloud computing has significantly increased the demand for high-performance data storage systems. A straightforward approach to provide a high performance storage system is using Solid-State Drives (SSDs). Inclusion of SSDs in storage systems, however, imposes significantly higher cost compared to Hard Disk Drives (HDDs). Recent studies suggest using SSDs as a caching layer for HDD-based storage subsystems in virtualization platforms. Such studies neglect to address the endurance and cost of SSDs, which can significantly affect the efficiency of I/O caching. Moreover, previous studies only configure the cache size to provide the required performance level for each VM, while neglecting other important parameters such as cache write policy and request type, which can adversely affect both performance-per-cost and endurance. In this paper, we present a new high-Endurance and Cost-efficient I/O Caching (ECI-Cache) scheme for virtualized platforms, which can significantly improve both the performance-per-cost and endurance of storage subsystems as opposed to previously proposed I/O caching schemes. Unlike traditional I/O caching schemes which allocate cache size only based on reuse distance of accesses, we propose a new metric, Useful Reuse Distance (URD), which considers the request type in reuse distance calculation, resulting in improved performance-per-cost and endurance for the SSD cache. Via online characterization of workloads and using URD, ECI-Cache partitions the SSD cache across VMs and is able to dynamically adjust the cache size and write policy for each VM. To evaluate the proposed scheme, we have implemented ECI-Cache in an open source hypervisor, QEMU (version 2.8.0), on a server running the CentOS 7 operating system (kernel version 3.10.0-327). Experimental results show that our proposed scheme improves the performance, performance-per-cost, and endurance of the SSD cache by 17%, 30% and 65%, respectively, compared to the state-of-the-art dynamic cache partitioning scheme.

이종 환경에서 대용량 RDF 데이터를 위한 동적 분할 기법

분산 환경에서 특정 서버에 발생한 부하 또는 서버 간의 통신으로 발생한 부하를 해결하기 위한 동적 분할이 필요하다. 이종 환경에서 기존 동적 분할 기법은 물리적 성능이 작은 서버에도 동일한 부하를 분배하므로 질의 응답시간이 늦어지는 문제가 발생한다. 본 논문에서는 이종 환경에서 대용량 RDF 데이터를 위한 동적 분할 기법을 제안한다. 제안하는 기법은 부하 분산을 위해 질의 빈도수와 질의에 사용된 정점 수를 가지고 질의 부하를 계산한다. 또한, 이종 환경에서 물리적 성능이 작은 서버에 적은 부하를 할당하기 위하여 서버들의 물리적 성능을 고려한 서버 부하를 계산한다. 부하 분산 시 서버 간의 통신량을 줄이기 위해 간선 절단 수가 최소가 되도록 동적 분할을 수행한다. 성능 평가를 통해 제안하는 동적분할 기법이 기존 동적 분할 기법에 비해 우수함을 입증한다.

Joint Resource Bidding and Tipping Strategies in Multi-Hop Cognitive Networks

In multi-hop secondary networks, bidding strategies for spectrum auction, route selection and relaying incentives should be jointly considered to establish multi-hop communication. In this paper, a framework for joint resource bidding and tipping is developed where users iteratively revise their strategies, which include bidding and incentivizing relays, to achieve their Quality of Service (QoS) requirements. A bidding language is designed to generalize secondary users' heterogeneous demands for multiple resources and willingness to pay. Then, group partitioning-based auction mechanisms are presented to exploit the heterogeneity of SU demands in multi-hop secondary networks. These mechanisms include primary operator (PO) strategies based on static and dynamic partition schemes combined with new payment mechanisms to obtain high revenue and fairly allocate the resources. The proposed auction schemes stimulate the participation of SUs and provide high revenue for the PO while maximizing the social welfare. Besides, they satisfy the properties of truthfulness, individual rationality and computational tractability. Simulation results have shown that for highly demanding users the static group scheme achieves 150% more winners and 3 times higher revenue for the PO compared to a scheme without grouping. For lowly demanding users, the PO may keep similar revenue with the dynamic scheme by lowering 50% the price per channel as the number of winners will increase proportionally.

Robust dynamic network traffic partitioning against malicious attacks

The continual growth of network traffic rates leads to heavy packet processing overheads, and a typical solution is to partition traffic into multiple network processors for parallel processing especially in emerging software-defined networks. This paper is thus motivated to propose a robust dynamic network traffic partitioning scheme to defend against malicious attacks. After introducing the conceptual framework of dynamic network traffic partitioning based on flow tables, we strengthen its TCP connection management by building a half-open connection separation mechanism to isolate false connections in the initial connection table (ICT). Then, the lookup performance of the ICT table is reinforced by applying counting bloom filters to cope with malicious behaviors such as SYN flooding attacks. Finally, we evaluate the performance of our proposed traffic partitioning scheme with real network traffic traces and simulated malicious traffic by experiments. Experimental results indicate that our proposed scheme outperforms the conventional ones in terms of packet distribution performance especially robustness against malicious attacks.

High-Endurance Hybrid Cache Design in CMP Architecture With Cache Partitioning and Access-Aware Policies

In recent years, nonvolatile memory (NVM) technologies, such as spin-transfer torque random-access memory (RAM) (STT-RAM) and phase change RAM, have drawn a lot of attention due to their low leakage and high density. However, both of these NVMs suffer from high write latency and limited endurance problems. To mitigate the write pressure on NVM, many static RAM (SRAM)/NVM hybrid cache designs have been proposed with write management policies. Unfortunately, existing hybrid cache designs do not consider the unbalanced workload of each core in (chip multiprocessor) architecture, resulting in unbalanced wear out of hybrid caches. This paper considers the unbalanced write distribution of a hybrid cache for CMP architecture as well as a novel hybrid cache design that includes SRAM cache, STT-RAM cache, and STT-RAM/SRAM hybrid cache banks. Based on the proposed hybrid cache design, two access-aware policies are proposed to mitigate unbalanced wearout of the STT-RAM region, and a wearout-aware dynamic cache partitioning scheme is proposed to dynamically partition the hybrid cache, improving the unbalanced write pressure among different cache partitions. The experimental results show that our proposed scheme and policies can achieve an average of 89 times improvement in cache lifetime and are able to reduce energy consumption by 58% compared with a SRAM cache.

Dynamic spatial partition density-based emergency message dissemination in VANETs

Location and density based emergency message broadcasting has attracted researchers attention in vehicular ad-hoc networks. However, most of current approaches do not provide good performance, in terms of delay in both light and dense traffic scenarios. Reliability in message delivery is another significant performance metric, especially in dense traffic scenarios. In this paper, we have analyzed and implemented a reliable time-efficient and multi-hop broadcasting scheme, called Dynamic Partitioning Scheme (DPS), which works well in both dense and light traffic scenarios. Our solid analytical evaluation and simulation results indicate that our proposed scheme outperforms five efficient broadcasting protocols in VANETs in terms of delay and reliability in emergency message broadcasting.

대규모 RDF 데이터의 분산 저장을 위한 동적 분할 기법

In recent years, RDF partitioning schemes have been studied for the effective distributed storage and management of large-scale RDF data. In this paper, we propose an RDF dynamic partitioning scheme to support load balancing in dynamic environments where the RDF data is continuously inserted and updated. The proposed scheme creates clusters and sub-clusters according to the frequency of the RDF data used by queries to set graph partitioning criteria. We partition the created clusters and sub-clusters by considering the workloads and data sizes for the servers. Therefore, we resolve the data concentration of a specific server, resulting from the continuous insertion and update of the RDF data, in such a way that the load is distributed among servers in dynamic environments. It is shown through performance evaluation that the proposed scheme significantly improves the query processing time over the existing scheme.

A Phase Behavior Aware Dynamic Cache Partitioning Scheme for CMPs

In multi-program environment, cache contention among processors can significantly degrade system performance. Cache partitioning served as an effective measure has been widely studied, especially for dynamic cache partitioning. However, it is difficult to decide the best cache quota which should be allocated to co-scheduled programs and the best time when a cache adjusting should be performed in dynamic cache partitioning scheme. This paper presents a novel dynamic cache partitioning mechanism based on the phase behavior of programs. It uses the performance monitoring units of modern processors and detects the phase behavior of programs to guide the cache partitioning at run-time. Since programs have recurring phase behavior during the whole execution time, on one hand, we can adjust the cache quota when a phase change occurs, on the other hand, we can make cache partitioning policy with higher accuracy and lower overhead by classifying phases. The method proposed in this work is validated in the measured results for applications from SPEC CPU 2006 benchmark suite. Compared with the performance of shared cache scheme, our method can achieve a speedup up to 1.214 for co-scheduled applications.

A new 3-D ray tracing method based on LTI using successive partitioning of cell interfaces and traveltime gradients

Abstract We present a new method of three-dimensional (3-D) seismic ray tracing, based on an improvement to the linear traveltime interpolation (LTI) ray tracing algorithm. This new technique involves two separate steps. The first involves a forward calculation based on the LTI method and the dynamic successive partitioning scheme, which is applied to calculate traveltimes on cell boundaries and assumes a wavefront that expands from the source to all grid nodes in the computational domain. We locate several dynamic successive partition points on a cell's surface, the traveltimes of which can be calculated by linear interpolation between the vertices of the cell's boundary. The second is a backward step that uses Fermat's principle and the fact that the ray path is always perpendicular to the wavefront and follows the negative traveltime gradient. In this process, the first-arriving ray path can be traced from the receiver to the source along the negative traveltime gradient, which can be calculated by reconstructing the continuous traveltime field with cubic B-spline interpolation. This new 3-D ray tracing method is compared with the LTI method and the shortest path method (SPM) through a number of numerical experiments. These comparisons show obvious improvements to computed traveltimes and ray paths, both in precision and computational efficiency.

PARALLEL SHOOTING AND BOUNCING RAY METHOD ON GPU CLUSTERS FOR ANALYSIS OF ELECTROMAGNETIC SCATTERING

This paper proposes an e-cient parallel shooting and bouncing ray (SBR) method on the graphics processing unit (GPU) cluster for solving the electromagnetic scattering problems. At each incident direction, the parallel SBR method partitions the virtual aperture into sub-apertures, and distributes the computational process of each sub-aperture over GPU nodes. As ray tubes in the virtual aperture do not have the same computational time, the parallel e-ciency highly depends on how to partition the virtual aperture. This paper addresses this issue by a dynamic partitioning scheme according to the computational time at the previous angle, which can achieve excellent load balance. Numerical examples are presented to demonstrate the accuracy, high parallel e-ciency, good scalability and versatility of the proposed method.

Dynamic cache partitioning based on hot page migration

Static cache partitioning can reduce inter-application cache interference and improve the composite performance of a cache-polluted application and a cache-sensitive application when they run on cores that share the last level cache in the same multi-core processor. In a virtualized system, since different applications might run on different virtual machines (VMs) in different time, it is inapplicable to partition the cache statically in advance. This paper proposes a dynamic cache partitioning scheme that makes use of hot page detection and page migration to improve the composite performance of co-hosted virtual machines dynamically according to prior knowledge of cache-sensitive applications. Experimental results show that the overhead of our page migration scheme is low, while in most cases, the composite performance is an improvement over free composition.

A branch-and-bound method for discretely-constrained mathematical programs with equilibrium constraints

We present a branch-and-bound algorithm for discretely-constrained mathematical programs with equilibrium constraints (DC-MPEC). This is a class of bilevel programs with an integer program in the upper-level and a complementarity problem in the lower-level. The algorithm builds on the work by Gabriel et al. (Journal of the Operational Research Society 61(9):1404–1419, 2010) and uses Benders decomposition to form a master problem and a subproblem. The new dynamic partition scheme that we present ensures that the algorithm converges to the global optimum. Partitioning is done to overcome the non-convexity of the Benders subproblem. In addition Lagrangean relaxation provides bounds that enable fathoming in the branching tree and warm-starting the Benders algorithm. Numerical tests show significantly reduced solution times compared to the original algorithm. When the lower level problem is stochastic our algorithm can easily be further decomposed using scenario decomposition. This is demonstrated on a realistic case.