Joint Task and Flow Scheduling for Time-Triggered and Strict-Priority Networks

Cluster computing imposes heavy demands on the communication network. Gigabit Ethernet technology can provide the required bandwidth to meet these demands. However, it has also shifted the communication bottleneck from network media to protocol processing. In this paper, we present an overview of Gigabit Ethernet technology and study the end-to-end Gigabit Ethernet communication bandwidth and latency. Performance graphs are collected using Net Pipe which clearly show the performance characteristics of TCP/IP over Gigabit Ethernet. These indicate the impact of a number of factors such as processor speeds, network adaptors, versions of the Linux Kernel and device drivers, and TCP/IP tuning on the performance of Gigabit Ethernet between two Pentium 11/350 PCs. Among the important conclusions are the marked superiority of the 2.1.121 and later development kernels and 2.2x production kernels of Linux and that the ability to increase the MTU beyond the Ethernet standard of 1500 could significantly enhance the throughput attainable.

Instability in packet-switching networks is normally associated with overload conditions, since queueing network models show that, in simple configurations, only overload generates instability. However, some results showing that instability can happen also in underloaded queueing networks began to appear about a decade ago. Underload instabilities can be produced by: 1) customer routes that visit the same queues several times; 2) variations of the customer service times at the different queues; and 3) complex scheduling algorithms. In this paper, we study, using fluid models and adversarial queueing theory, possible underload instabilities due to flow schedulers in packet networks, focusing on output queued switches with strict priority (SP) schedulers and Generalized Processor Sharing (GPS) schedulers. The considered scenarios always refer to acyclic packet routes and consider customer service times that vary only according to channel capacities, thus resembling the approaches being currently considered to provide QoS in the Internet. Our (in)stability results are rather surprising: SP schedulers appear to be more robust than GPS schedulers whenever exact information on the effective average packet flow rates is not available.

While an SSD (Solid State Drive) has been widely used for storage in many computing systems due to its small average latency, how to provide timing guarantees of a delay-sensitive (real-time) task on a real-time system equipped with an SSD has not been fully explored. A recent study has proposed a work-constraining I/O scheduler, called K2, which has succeeded in reducing the tail latency of a real-time task at the expense of compromising the total bandwidth for real-time and non-real-time tasks. Although the queue length bound parameter of the K2 scheduler is a key to regulate the tradeoff between a decrease in the tail latency of a real-time task and an increase in penalty of the total bandwidth, the parameter’s impact on the tradeoff has not been thoroughly investigated. In particular, no studies have addressed how the case of a fully occupied SSD that incurs garbage collection changes the performance of the K2 scheduler in terms of the tail latency of the real-time task and the total bandwidth. In this paper, we systematically analyze the performance of the K2 scheduler for different I/O operation types, based on experiments on Linux. We investigate how the performance is changed on a fully occupied SSD due to garbage collection. Utilizing the investigation, we draw general guidelines on how to select a proper setting of the queue length bound for better performance. Finally, we propose how to apply the guidelines to achieve target objectives that optimize the tail latency of the real-time task and the total bandwidth at the same time, which has not been achieved by previous studies.

Network Calculus is an established analytical framework which can provide deterministic estimates of latency and buffer requirements for feed-forward packet-switched communication networks. Researchers find it useful in analyzing network performance in real-time systems. As there are several Network Calculus tools and software packages for estimating network performance, it is important to select suitable Network Calculus tools for network analysis problems. This article introduces twelve established tools for Network Calculus analysis and compares their advantages and disadvantages in terms of their applicability for various systems and analysis needs. We also provide recommendations for users to select the suitable tools for research. Five public tools, namely DiscoDNC, RTC Toolbox, Deborah, CyNC, and nc-tandem-tight, are chosen to analyze end-to-end delay in a series of tandem networks for their performance demonstration. Simulations of the same tandem networks are implemented to compare with the analytical results achieved by the Network Calculus tools. To our best knowledge, this is the first comprehensive study that compares existing tools of Network Calculus analysis for real-time system network infrastructure. Our aim is to provide quick guidance to research communities for selecting suitable tools to analyze delay performance of networks.

This paper reports recent work on gigabit Ethernet interconnection communication traffic analysis. We proposed and demonstrated the system concept of gigabit Ethernet communication. 1.25Gbps, 10Gpbs and 40Gpbs gigabit Ethernet interconnection networks are considered for computer communications. Various switching structures, such as crossbar, double crossbar (Dcrossbar), Modified Dilated Benes (MDB), General MDB (GMDB), Benes, Dilated Benes (Dbenes), Tree architecture, Simplified tree (Stree), Extended baseline network (Ebaseline) are analyzed for searching the optimal performance of gigabit Ethernet communication. The numerical results for computation information transferring can be applied to search a best strategy for gigabit Ethernet communication networks interconnections.

The terrestrial point-to-multipoint fixed wireless access (PMP FWA) systems have evolved from their first circuit-switched proprietary implementations in the 1970s into flexible, packet-based solutions available to competitive local exchange carriers (CLEC) for providing POTS and Internet access in WAN and MAN. Examples of such systems can be seen in a number of proprietary technologies, as well as in the commercial implementations of the IEEE 802 families of standards. The purpose of this paper is to evaluate the performance of a PMP FWA system in the context of today's packet based IntServ infrastructure. The evaluation takes into account three scheduling algorithms (FIFO, strict priority, and lead-time) in a number of different configurations. This paper also proposes a set of mechanisms that can be introduced into the IntServ framework to support the lead-time scheduling algorithm.

IEEE 802.16 well known as WIMAX is defined as the standard of Broadband WMAN networks. This technology is able to mix real time and non real time services with the definition of service classes. To efficientlty schedule traffics of different classes, we proposed in a previous work a token bucket based scheme and evaluated its performances by means of probabilistic models. In this paper, we explore the use of TPN (Time Petri nets) for deterministic performance evaluation in order to compute quantitative and qualitative properties. We show how to model two different strategies of traffic scheduling by using TPN model extended to logical inhibitor arcs. The first strategy is based on strict priorities between traffics, the second is based on token bucket mechanism. Then in order to improve the efficiency of the latter scheme, we propose to extend it to a QoS-management mechanism. Then we simulate the obtained models by using the tool TINA in order to evaluate their performances when varying the load of the system. Preliminary results show that token bucket extension provides better figures in terms of service delay, than other schemes, in case of high service loads while maintaining fairness and priorities between traffics.

In the context of Edge Computing (EC) and Internet of Things (IoT), numerous tasks are offloaded from mobile users and sensor devices to edge nodes for further processing to reduce delay and solve the problem of insufficient local computation resources. These tasks can be mainly divided into delay-sensitive and computation-oriented tasks. The former tasks depend on the service provided by the container, while the latter tasks are submitted as a batch with task dependencies. Considering the heterogeneity of edge nodes, joint task scheduling can effectively improve resource utilization. However, relatively few researches consider the different characteristics of tasks like container constraints and task dependencies in joint task scheduling in EC. In order to fill in this gap, we propose a deep deterministic policy gradient (DDPG) based online joint task scheduling (OJTS) algorithm. Specifically, 1) We first model the problem of joint scheduling of delay-sensitive and computation-oriented tasks in resource-constrained EC scenario with the goals of maximizing system utility and minimizing system cost (weighted sum of the number and duration of unfinished tasks). 2) Then, we propose a deep reinforcement learning (DRL) algorithm to solve the above problem and make appropriate adjustments to the original network structure according to the scheduling decision. 3) Through validation on real-world trace, OJTS can improve the system utility by 26.0% and overall reward by 51.2% compared with baselines and meet real-time decision-making requirements.

This paper presents a performance study of a nonrigid registration algorithm for investigating lung disease on clusters. Our algorithm combines two conventional acceleration techniques in order to achieve fast registration: a data-parallel processing technique for accelerating the registration procedure; and a precomputation technique for reducing the computational complexity. We perform some experiments on three clusters with different CPU and network performance in order to make clear what kinds of acceleration techniques and computing environments provide higher performance. The results show that a cluster with Gigabit Ethernet (GbE) network is the most cost effective solution that reduces registration time from ten hours to ten minutes with a linear speedup.

: This report documents the installation, acceptance testing, vendor interactions, and final system acceptance for production usage of a dedicated high performance computing project investment that was installed at the U.S. Army Research Laboratory facilities on the Aberdeen Proving Ground in MD. The uniqueness of this system lies in the use of the low latency 10 GigE (gigabit Ethernet) network fabric, many-core AMD Opteron central processing units (CPUs), Nvidia Fermi graphics processing unties (GPUs), and a mixture of CPU-only and CPU/GPU nodes. The final accepted production system is stable and capable of nearly 1 PetaFLOPS (1012 floating-point operations per second) of single precision performance.

With the advent of Gigabit Ethernet network technology, a new contender for the next-generation cluster interconnect is on the horizon. We benchmark Gigabit Ethernet and compare it to other cluster interconnects, namely Fast Ethernet, Myrinet and Scalable Coherent Interface. For a meaningful comparison, benchmark experiments are carried out at two levels: TCP/IP networking and MPI parallel programming. Using high-end PCs (Pentium III 450 MHz) and standard system software (Linux and MPICH), our results show that with Gigabit Ethernet, end-to-end throughputs of up to 44 MBytes per second can be achieved. A change from Fast Ethernet to Gigabit Ethernet resulted in performance gains of up to a factor of 5.05 for ftp, up to a factor of 3.10 for MPI point-to-point communication and up to a factor of 2.58 for some NAS parallel benchmarks, depending on communication granularity and message size. Further performance gains seem to require improved protocol processing and node hardware (memory bandwidth, file system, PCI bus).

Existing scheduling and decision solutions to container terminal logistics systems (CTLS) are falling short of peoples' wishes both in theory and in practice. This paper proposes a container terminal gang scheduling computational paradigm (GSCP) for planning and control at container terminals based on computational thinking. GSCP is built on the computing perspective and architecture of multi-processor system-on-chip. GSCP applies principles of pervasive computing to evaluate load conditions of core resources, and then integrates the fundamental principles of gang scheduling, processor affinity and load balancing to define the flexible and robust decision framework and scheduling algorithm set. Those are intended to improve the performance of CTLS, and strive for balance among through capacity for vessels and containers, task latency and load leveling as well. GSCP is demonstrated and validated by a typical container hub logistics service case with intensive computational experiments.

SoCs for multimedia applications typically use only one port to off-chip DRAM for cost reasons. The sharing of interconnect and the off-chip DRAM port by several IP blocks makes the performance of a SoC under design hard to predict. Network calculus defines the concept of flow and has been successfully used to analyse the performance of communication networks. We propose to apply network calculus to the verification of memory access latencies. Two novel network elements, packet stretcher and packet compressor, are used to model the SoC interconnect and DRAM controller. We further extend the flow concept with a degree and make use of the peak characteristics of a flow to tighten the bounds in the analysis. We present a video playback case study and show that the proposed application of network calculus allows us to statically verify that all requirements on memory access latency are fulfilled.

Summary form only given. We investigate the problem of propagating of data to nodes and scheduling of available nodes in a dynamic data communication network. The goal of scheduling algorithm is to find a request-service scheduling such that the total communication cost is minimized. In the special case in which all requests are of equal priorities and all services have equal performance, the scheduling problem becomes the mapping of the maximum number of requests to the free service-nodes. Scheduling in a communication network entails the efficient search of a mapping from request-nodes to free service-nodes such that circuit blockage in the network are minimized and service-nodes are maximally utilized.

This paper presents a smart monitoring technique for photovoltaic power systems by using wire and wireless communication networks in which the RS-232/484 and the Zigbee communication networks are inherently established respectively. In the proposed monitoring systems, environmental data sequences and the output power measured by sensors in photovoltaic systems are transferred to PC systems via two communication networks. We made electronic hardware boards for sensors and communication networks to construct its real-time monitoring system and carry out experiments for demonstrating reliability of the proposed monitoring system.