A framework for QoS-aware software components

SummaryThe execution of agent‐based simulations (ABSs) on hardware accelerator devices such as graphics processing units (GPUs) has been shown to offer great performance potentials. However, in heterogeneous hardware environments, it can become increasingly difficult to find viable partitions of the simulation and provide implementations for different hardware devices. To automate this process, we present OpenABLext, an extension to OpenABL, a model specification language for ABSs. By providing a device‐aware OpenCL backend, OpenABLext enables the co‐execution of ABS on heterogeneous hardware platforms consisting of central processing units, GPUs, and field programmable gate arrays (FPGAs). We present a novel online dispatching method that efficiently profiles partitions of the simulation during run‐time to optimize the hardware assignment while using the profiling results to advance the simulation itself. In addition, OpenABLext features automated conflict resolution based on user‐specified rules, supports graph‐based simulation spaces, and utilizes an efficient neighbor search algorithm. We show the improved performance of OpenABLext and demonstrate the potential of FPGAs in the context of ABS. We illustrate how co‐execution can be used to further lower execution times. OpenABLext can be seen as an enabler to tap the computing power of heterogeneous hardware platforms for ABS.

A new information integration platform for computer integrated manufacturing system (CIMS) is presented, which is based on agent and CORBA. CORBA enhances the system integration because it is an industry-standard for interoperable, distributed objects across heterogeneous hardware and software platform. Agent technology is used to improve intelligence of integration system. In order to implement the information integration platform, we use network integration server to integrate network, design a generic database agent to integrate database, adopt multi-agent based architecture to integrate application, and utilize wrapper as CORBA object to integrate legacy code.

A city natural gas storage station control system is presented in this paper. It consists of a Station Control Center, a process control station used Rockwell PLC, two desiccators control station used Ormon PLC, three high pressure compressors control station used Siemens PLC, Ethernet communication between supervise control computer and Rockwell PLC, RS 485 communication among the Siemens and Ormon PLC. Based on object character analysis and application requirement discussion, normal functions have realized in this system, such as flow accumulation at PLC, timestamp of data, and real-time control. The OPC technology is adopted to accomplish the vertical integration among the heterogeneous network, heterogeneous system, and heterogeneous platform. Data can be reused in the opening system. The remote scheduling control is realized at a scheduling and supervise control center which is 42 kilometers away from Wenxingwan station. This solution can be applied to more kind of trades such as oil, power transmission, water supply, and so on.

Admission control plays a very important role in wireless systems, as it is one of the basic mechanisms for ensuring the quality of service offered to users. Based on the available network resources, it estimates the impact of adding or dropping a new session request. In both 2G and 3G systems, admission control refers to a single network. As we are moving towards heterogeneous wireless networks referred to as systems beyond 3G or 4G, admission control will need to deal with many heterogeneous networks and admit new sessions to a network that is most appropriate to supply the requested QoS. In this article we present the fundamentals of access-network-based admission control, an overview of the existing admission control algorithms for 2G and 3G networks, and finally give the design of a new admission control algorithm suitable for future 4G networks and specifically influenced by the objectives of the European WINNER project.

Multi-tenant inference, as a prevalent inference paradigm nowadays, requires deploying multiple deep learning models on the hardware platform to concurrently process inference tasks. Modern platforms are typically equipped with various heterogeneous processors, such as CPU-GPU platform. To reduce resource contention and improve Quality of Service (QoS) in the multi-tenant scenario, existing work has studied cross-processor inference at the model- and layer-level. However, coarse-grained scheduling cannot flexibly account for subtle resource fluctuations, which may lead to task blockages and incur significant processor switching overheads. Such work usually requires extensive modification and retraining of the models. Therefore, we propose a finer-grained operator-level cross-processor scheduling framework COS, which can more precisely divide the computational workloads and switching overheads for the tenants, without modifying or retraining. We introduce a novel intermediate representation to abstract and simplify the scheduling problem, and propose an efficient two-phase search algorithm. COS is automated and easy-to-scale, through experiments on various heterogeneous hardware platforms and models, we demonstrate that COS is more flexible and effective than layer-level scheduling, and achieves higher throughput than single-processor processing in the multi-tenant scenario. Furthermore, COS is an offline optimization method, and its overhead is highly acceptable.

Portable virtual cycle accounting for large-scale distributed cycle sharing systems

Many application domains are increasingly leveraging service-oriented architecture (SOA) techniques to facilitate rapid system deployment. Many of these applications are time-critical and, hence, real-time assurance is an essential step in the service composition process. However, there are gaps in existing service composition techniques for real-time systems. First, admission control is an essential technique to assure the time bound for service execution, but most of the service composition techniques for real-time systems do not take admission control into account. A service may be selected for a workflow during the composition phase, but then during the grounding phase, the concrete service may not be able to admit the workload. Thus, the entire composition process may have to be repeated. Second, communication time is an important factor in real-time SOA, but most of the existing works do not consider how to obtain the communication latencies between services during the composition phase. It is clear that maintaining a full table of communication latencies for all pairs of services is infeasible. Obtaining communication latencies between candidate services during the composition phase can also be costly, since many candidate services may not be used for grounding. Thus, some mechanism is needed for estimating the communication latency for composite services. In this paper, we propose a three-phase composition approach to address the above issues. In this approach, we first use a highly efficient but moderately accurate algorithm to eliminate most of the candidate compositions based on estimated communication latencies and assured service response latency. Then, a more accurate timing prediction is performed on a small number of selected compositions in the second phase based on confirmed admission and actual communication latency. In the third phase, specific concrete services are selected for grounding, and admissions are actually performed. The approach is scalable and can effectively achieve service composition for satisfying real-time requirements. Experimental studies show that the three-phase approach does improve the effectiveness and time for service composition in SOA real-time systems. In order to support the new composition approach, it is necessary to effectively specify the needed information. In this paper, we also present the specification model for timing-related information and the extension of OWL-S to support this specification model.

Service-Oriented Architectures (SOAs) are increasingly being used for designing and building large-scale networked and distributed systems. Catering to the complex and dynamically varying needs of business applications/clients, these systems must usually be realized by dynamically composing a variety of network-available services. Evaluation of large-scale SOAs, particularly on dynamic network platforms, such as Mobile Ad-hoc Networks (MANETs), is a non-trivial problem that requires not only a correct modeling of SOAs and the network platform, but also their relationships. This paper describes a new tool – SOAMANET – to design and rapidly synthesize simulations for the experimental evaluation of SOAs on MANET platforms. With its modeling techniques and analysis capabilities, SOAMANET allows simulation-based and system execution-based analysis of dynamic SOA and/or MANET designs and implementations.

Integration of Networked Control Systems is always an engineering challenge. Heterogeneous hardware and software environments combined with long lifetime of control systems makes this job particularly troublesome. Modern concepts of integration of automation systems are related to Service Oriented Architecture (SOA). While its suitability is proven in IT systems, SOA has not been adopted yet in commercial Programmable Logic Controllers (PLC), and thus cannot be considered as a solution for integration with already deployed control systems. However, during past years thousands of PLCs with embedded HTTP servers were deployed in the field. These devices, used together with modern PLC that acts as the HTTP client, enable unique opportunity of integration for control systems with soft real-time constraints. In the present study, the performance of PLC-to-PLC communications based on HTTP is evaluated and compared to Modbus TCP.

We study the integration of terrestrial-satellite networks through the introduction of QoS architectures inside the two networks. We introduce integrated services architecture in the satellite segment and differentiated services architecture in the terrestrial network. We examine the admission control of the integrated terrestrial-satellite network. We propose an admission control which consists of two phases: terrestrial admission control and satellite admission control. We show two approaches relative to the terrestrial call admission control based on aggregate traffic control: a more conservative DPS (dynamic packet state) approach; a more scalable ER (egress/gateway) approach. Different simulations are outlined and the conformance to QoS parameters for both approaches has been verified. Furthermore, we have verified that the ER approach leads to a performance improvement of the overall system in terms of calls accepted on the terrestrial segment and in terms of more reduced control messages than the DPS approach.

Nowadays wireless sensor netwoks (WSN) technology, wireless communications and digital electronics have made it realistic to produce a large scale miniaturized devices integrating sensing, processing and communication capabilities. The focus of this paper is to present an innovative mobile platform for heterogeneous sensor networks, combined with adaptive methods to optimize the communication architecture for novel potential applications in multimedia and entertainment. In fact, in the near future, some of the applications foreseen for WSNs will employ multi-platform systems with a high number of different devices, which may be completely different in nature, size, computational and energy capabilities, etc. Nowadays, in addition, data collection could be performed by UAV platforms which can be a sink for ground sensors layer, acting essentially as a mobile gateway. In order to maximize the system performances and the network lifespan, the authors propose a recently developed hybrid technique based on evolutionary algorithms. The goal of this procedure is to optimize the communication energy consumption in WSN by selecting the optimal multi-hop routing schemes, with a suitable hybridization of different routing criteria. The proposed approach can be potentially extended and applied to ongoing research projects focused on UAV-based sensing with WSN augmentation and real-time processing for immersive media experiences.

In this paper, we propose a new admission control scheme for the VBR MPEG video in the VOD system over ATM-over-DSL access networks. In the proposed scheme, when a new video call is requested, the video server decides if the VOD system can provide the video call with its negotiated QoS without violating the QoS guarantees of existing calls in the network.

An unmanned swarm is usually composed of a group of homogeneous or heterogeneous hardware platforms, software control systems, and interfaces for human–computer interaction that operate collectively to achieve a specific goal by information interaction. They exhibit robustness and fault tolerance when facing complex missions, making it crucial in military, transportation, intelligent traffic, and other fields. However, the coupling between the hardware and software of a heterogeneous unmanned swarm can indeed have significant implications for system flexibility, software development and deployment, and hardware maintenance. Over the years, there has been a significant shift from traditional hardware-focused control systems to a greater emphasis on the core software layer. In this paper, a distributed network architecture is proposed to solve this problem, in which hardware resources are abstracted and represented to accomplish standardization and unification by defining a consistent and uniform set of data formats, and a resource pool of hardware data is constructed to realize the function that the number and scale of platforms is irrelevant, the task module can be plug-and-play at any time, and the software can be configured on demand. The resource scheduling of a single platform is achieved through process and thread communication using shared memory, while the resource scheduling of a cross platform is achieved through a network using request and response and subscription and notification. As a result, it can satisfy the development of functional modules in a software-defined mode and gradually improve the intelligence capability of an unmanned swarm. Based on the above architecture, the overall framework of the autonomous navigation system and the collaborative control system has been successfully established. Finally, a hardware-in-the-loop simulation environment is constructed, and the integration and verification of the proposed distributed architecture is carried out by the cooperative formation experiment, which proves the feasibility of this proposal.

Heterogeneous hardware platforms are often used for implementing complex critical real-time applications, like Advanced driver-assistance systems (ADAS) and autonomous driving. Typically, they are composed of CPU hosts and a set of accelerators. To better support real-time workloads, several hardware accelerators have evolved to allow preemption for computationally intensive tasks, such as GPUs. However, their preemption costs can be very high compared to classical CPU preemption, and therefore must be taken into account at design time and in the scheduling analysis. In this paper, we address mainly two tightly correlated problems: (i) task allocation for a set of real-time tasks, modeled by conditional directed acyclic graphs (C-DAG), onto multiprocessor platforms under partitioned preemptive Earliest Deadline First scheduling, assuming a non-negligible cost of preemption, and (ii) intermediate deadlines and offsets assignments to real-time C-DAGs, so to remove unnecessary preemption and reduce the total preemption overhead. The effectiveness of the proposed technique is evaluated using a large set of synthetic tasks sets.

Deploying Convolutional Neural Networks (CNNs) efficiently across heterogeneous hardware platforms—such as CPUs, GPUs, and dedicated AI accelerators—remains a significant challenge due to differences in memory bandwidth, compute units, software stacks, and kernel implementations. While Network Architecture Search (NAS) has produced highly compute-efficient architectures like MobileNet, many of these models underperform in real deployments because they optimize primarily for FLOPs/MACs rather than system-level constraints. In this paper, we present a hardware-aware approach to architecture evaluation using the Once-for-All (OFA) framework, which enables flexible exploration of sub-networks with varying input resolution, depths, widths and kernel sizes. We conduct a detailed analysis of widely popular architecture like MobileNet under varying hyperparameters—including input resolution, depth, width and kernel size —and evaluate their behavior across CPUs, GPUs, and AI accelerators. By correlating model structure with real execution performance, we provide actionable insights that help developers design architectures that better exploit hardware capabilities. Our approach integrates system-level performance evaluation with guidelines that bridge the gap between theoretical model efficiency and practical deployment performance, enabling more predictable execution on modern AI hardware.