Scheduling Configuration Research Articles

ITune: Engineering the Performance of Xen Hypervisor via Autonomous and Dynamic Scheduler Reconfiguration

Despite the widespread use of server virtualization technologies in cloud data centers, system administrators experience multiple challenges in configuring the hypervisor’s scheduler parameters to optimize its performance. Manually tuning the scheduler’s parameters is a common practice, however, this approach is not effective particularly when dealing with dynamically changing workload and resource utilizations on the host machines. This problem becomes even harder if cloud resources are overbooked while hosting both latency-sensitive and batched applications. To address these issues, this paper presents iTune, which is a framework for engineering the performance of a hypervisor intelligently via autonomous scheduler configurations. Concretely, iTune optimizes the Xen hypervisor’s scheduler configuration parameters autonomously through a three phase process comprising: (1) Discoverer, which monitors and saves the resource usage history of the host machines and groups set of related host machine workloads, (2) Optimizer, where optimum Xen scheduler configuration parameters for each workload cluster are explored by employing a simulated annealing machine learning algorithm, and (3) Observer, where iTune monitors the resource usage of host machines online, classifies them into one of the categories found in the Discoverer phase, and loads the optimum scheduler parameters determined in the Optimizer phase. Experimental results validate our claims.

Comparative Simulation Study of Production Scheduling in the Hybrid and the Parallel Flow

AbstractScheduling is one of the most important decisions in production control. An approach is proposed for supporting users to solve scheduling problems, by choosing the combination of physical manufacturing system configuration and the material handling system settings. The approach considers two alternative manufacturing scheduling configurations in a two stage product oriented manufacturing system, exploring the hybrid flow shop (HFS) and the parallel flow shop (PFS) environments. For illustrating the application of the proposed approach an industrial case from the automotive components industry is studied. The main aim of this research to compare results of study of production scheduling in the hybrid and the parallel flow, taking into account the makespan minimization criterion. Thus the HFS and the PFS performance is compared and analyzed, mainly in terms of the makespan, as the transportation times vary. The study shows that the performance HFS is clearly better when the work stations’ processing times are unbalanced, either in nature or as a consequence of the addition of transport times just to one of the work station processing time but loses advantage, becoming worse than the performance of the PFS configuration when the work stations’ processing times are balanced, either in nature or as a consequence of the addition of transport times added on the work stations’ processing times. This means that physical layout configurations along with the way transport time are including the work stations’ processing times should be carefully taken into consideration due to its influence on the performance reached by both HFS and PFS configurations.

Semi-Supervised Clustering Algorithms for Grouping Scientific Articles

Creating sessions in scientific conferences consists in grouping papers with common topics taking into account the size restrictions imposed by the conference schedule. Therefore, this problem can be considered as semi-supervised clustering of documents based on their content. This paper aims to propose modifications in traditional clustering algorithms to incorporate size constraints in each cluster. Specifically, two new algorithms are proposed to semi-supervised clustering, based on: binary integer linear programming with cannot-link constraints and a variation of the K-Medoids algorithm, respectively. The applicability of the proposed semi-supervised clustering methods is illustrated by addressing the problem of automatic configuration of conference schedules by clustering articles by similarity. We include experiments, applying the new techniques, over real conferences datasets: ICMLA-2014, AAAI-2013 and AAAI-2014. The results of these experiments show that the new methods are able to solve practical and real problems.

Throughput Optimization of Continuous Biopharmaceutical Manufacturing Facilities.

In order to operate profitably under different product demand scenarios, biopharmaceutical companies must design their facilities with mass output flexibility in mind. Traditional biologics manufacturing technologies pose operational challenges in this regard due to their high costs and slow equipment turnaround times, restricting the types of products and mass quantities that can be processed. Modern plant design, however, has facilitated the development of lean and efficient bioprocessing facilities through footprint reduction and adoption of disposable and continuous manufacturing technologies. These development efforts have proven to be crucial in seeking to drastically reduce the high costs typically associated with the manufacturing of recombinant proteins. In this work, mathematical modeling is used to optimize annual production schedules for a single-product commercial facility operating with a continuous upstream and discrete batch downstream platform. Utilizing cell culture duration and volumetric productivity as process variables in the model, and annual plant throughput as the optimization objective, 3-D surface plots are created to understand the effect of process and facility design on expected mass output. The model shows that once a plant has been fully debottlenecked it is capable of processing well over a metric ton of product per year. Moreover, the analysis helped to uncover a major limiting constraint on plant performance, the stability of the neutralized viral inactivated pool, which may indicate that this should be a focus of attention during future process development efforts.LAY ABSTRACT: Biopharmaceutical process modeling can be used to design and optimize manufacturing facilities and help companies achieve a predetermined set of goals. One way to perform optimization is by making the most efficient use of process equipment in order to minimize the expenditure of capital, labor and plant resources. To that end, this paper introduces a novel mathematical algorithm used to determine the most optimal equipment scheduling configuration that maximizes the mass output for a facility producing a single product. The paper also illustrates how different scheduling arrangements can have a profound impact on the availability of plant resources, and identifies limiting constraints on the plant design. In addition, simulation data is presented using visualization techniques that aid in the interpretation of the scientific concepts discussed.

GA-ETI: An enhanced genetic algorithm for the scheduling of scientific workflows in cloud environments

Over recent years, cloud computing has become one of the main sources of computer power to run scientific experiments. To cope with these demands, cloud providers need to efficiently match applications with computing resources to maintain an acceptable level of customer satisfaction. A correct match or scheduling of scientific workflows relies on the ability to fully analyze applications prior to execution, analyze characteristics of available computing resources, provide users with several scheduling configurations, and guide users to select the optimal configuration to execute workflows. To date, different schedulers have been proposed to execute complex applications on cloud environments; nevertheless, none exists, to the best of our knowledge, to provide all the aforementioned features. GA-ETI, the scheduler proposed in this work, is designed to address all aforementioned concerns by providing several efficient solutions (in a Pareto Front fashion) to run scientific workflows on cloud resources. Flexibility of optimization procedure of GA-ETI allows it to easily adapt to different types of scientific workflows and produce schedules that effectively exploit/consider the relationship between jobs and their required data. GA-ETI acts as an interface between cloud user and cloud provider in receiving an application, analyzing it, and distributing its tasks among selected resources. GA-ETI differs from the majority of proposed schedulers because it can adapt to the size of both jobs and virtual machines, it includes a monetary cost model (from a public cloud), and it considers complex interdependencies among tasks. We test GA-ETI with five well-known benchmarks with different computing and data transfer demands in our VMware-vSphere private cloud. Through experimentation, GA-ETI has been proved to reduce makespan of executing workflows between 11% and 85% when compared to three up-do-date scheduling algorithms without increasing the monetary cost. GA-ETI opens the way to develop a top-layer-scheduler for a workflow manager system to provide a complex analysis and include different optimizing objectives.

Optimal radiotherapy dose schedules under parametric uncertainty

We consider the effects of parameter uncertainty on the optimal radiation schedule in the context of the linear-quadratic model. Our interest arises from the observation that if inter-patient variability in normal and tumor tissue radiosensitivity or sparing factor of the organs-at-risk (OAR) are not accounted for during radiation scheduling, the performance of the therapy may be strongly degraded or the OAR may receive a substantially larger dose than the allowable threshold. This paper proposes a stochastic radiation scheduling concept to incorporate inter-patient variability into the scheduling optimization problem. Our method is based on a probabilistic approach, where the model parameters are given by a set of random variables. Our probabilistic formulation ensures that our constraints are satisfied with a given probability, and that our objective function achieves a desired level with a stated probability. We used a variable transformation to reduce the resulting optimization problem to two dimensions. We showed that the optimal solution lies on the boundary of the feasible region and we implemented a branch and bound algorithm to find the global optimal solution. We demonstrated how the configuration of optimal schedules in the presence of uncertainty compares to optimal schedules in the absence of uncertainty (conventional schedule). We observed that in order to protect against the possibility of the model parameters falling into a region where the conventional schedule is no longer feasible, it is required to avoid extremal solutions, i.e. a single large dose or very large total dose delivered over a long period. Finally, we performed numerical experiments in the setting of head and neck tumors including several normal tissues to reveal the effect of parameter uncertainty on optimal schedules and to evaluate the sensitivity of the solutions to the choice of key model parameters.

Workload Known VMM Scheduler for Server Consolidation for Enterprise Cloud Data Center

This study proposes a novel adaptive meta-heuristics based scheduling policies for provisioning the VCPU resources among competing VM service domains in a cloud. Such provisioning guarantees to Service Level Agreement for each domain, with respect to the diverse workloads on-the-fly. The framework is built on CSIM models and tools, making it easy to understand and configure various virtualization setups. The study demonstrates the usefulness of the framework by evaluating proactive, reactive and adaptive VCPU scheduling algorithms. The paper evaluates how periodic/aperiodic execution of control actions can affect policy performance and speed of convergence. The periodic reactive resource allocation is used as the baseline for analysis and the average response time is the performance metric. Simulation based experiments using variety of real-world arrival traces and synthetic workloads results that the proposed provisioning technique detects changes in arrival pattern and resource demands and allocates resources accordingly to achieve application SLA targets. The proposed model improves CPU utilization and makes the best tradeoff between resource utilization and performance from 2 to 6% comparing with the default VMM scheduler configurations for diverse workloads. In addition, the results of the experiments show that the proposed Weighed Moving Average algorithm combined with the aperiodic policy significantly outperforms other dynamic VM consolidation algorithms in all cases, in regard to the SLA metric due to a substantially reduced level of response time violations and the frequency of algorithm invocation.

Modeling Message Queueing Services with Reliability Guarantee in Cloud Computing Environment Using Colored Petri Nets

Motivated by the need for loosely coupled and asynchronous dissemination of information, message queues are widely used in large-scale application areas. With the advent of virtualization technology, cloud-based message queueing services (CMQSs) with distributed computing and storage are widely adopted to improve availability, scalability, and reliability; however, a critical issue is its performance and the quality of service (QoS). While numerous approaches evaluating system performance are available, there is no modeling approach for estimating and analyzing the performance of CMQSs. In this paper, we employ both the analytical and simulation modeling to address the performance of CMQSs with reliability guarantee. We present a visibility-based modeling approach (VMA) for simulation model using colored Petri nets (CPN). Our model incorporates the important features of message queueing services in the cloud such as replication, message consistency, resource virtualization, and especially the mechanism named visibility timeout which is adopted in the services to guarantee system reliability. Finally, we evaluate our model through different experiments under varied scenarios to obtain important performance metrics such as total message delivery time, waiting number, and components utilization. Our results reveal considerable insights into resource scheduling and system configuration for service providers to estimate and gain performance optimization.

Deployment of a WLCG network monitoring infrastructure based on the perfSONAR-PS technology

The WLCG infrastructure moved from a very rigid network topology, based on the MONARC model, to a more relaxed system, where data movement between regions or countries does not necessarily need to involve T1 centres. While this evolution brought obvious advantages, especially in terms of flexibility for the LHC experiment's data management systems, it also opened the question of how to monitor the increasing number of possible network paths, in order to provide a global reliable network service. The perfSONAR network monitoring system has been evaluated and agreed as a proper solution to cover the WLCG network monitoring use cases: it allows WLCG to plan and execute latency and bandwidth tests between any instrumented endpoint through a central scheduling configuration, it allows archiving of the metrics in a local database, it provides a programmatic and a web based interface exposing the tests results; it also provides a graphical interface for remote management operations. In this contribution we will present our activity to deploy a perfSONAR based network monitoring infrastructure, in the scope of the WLCG Operations Coordination initiative: we will motivate the main choices we agreed in terms of configuration and management, describe the additional tools we developed to complement the standard packages and present the status of the deployment, together with the possible future evolution.

MAPS: Mapping Concurrent Dataflow Applications to Heterogeneous MPSoCs

Processor Systems on Chip (MPSoCs) in order to cope with the increasing applications demands and the tight energy budget of portable devices. The complexity of these systems makes them difficult to program, which has caused academia and industry to look for alternative methodologies and models. In the particular case of multimedia and baseband processing, dataflow models are being proposed and appear to be a sensible choice to represent applications. While high-level models, like dataflow, increase programmers' productivity, new, powerful tools are badly required that lower the abstract specification into an efficient implementation. In this paper, a framework is presented that provides support for mapping multiple dataflow applications onto heterogeneous MPSoCs. The framework is aware of design constraints, provides different means for performance estimation and supports a variety of mapping heuristics. The tool is showcased on three applications on a virtual platform containing heterogeneous processing elements. The heuristics for single applications reported a speedup of up to 40% when compared against random walk. The multi-application component helped to find an appropriate scheduling configuration that met real-time constraints when the three applications were running simultaneously.

Optimal Converter Aisle Scheduling in a Nickel Smelting Plant

The manual scheduling of operations in a Converter Aisle for Nickel Smelting is both a time-consuming and tedious task. The challenge of manually scheduling this process involves several tasks on multiple units (such as tapping of matte from furnaces, charging of converters, blowing, and skimming slag from converters). Furthermore, these tasks need to be carried out subject to a number of operational constraints that include upper and lower matte level limits in the furnaces, and the number and timing of the blowing operations required for a converter batch. This already complex set of tasks is further complicated by emission limits, which places additional constraints on the Converter Aisle operation. However, when process disruptions invalidate the nominal schedule, it falls on operators to rely on operating protocols and experience to navigate the process operations and obey operating constraints as best as possible. These considerations motivate the development and use of an optimal scheduling decision support system that is capable of timely process scheduling, while respecting operational constraints, to achieve a given objective. The work presented formulates the operation scheduling as a mathematical optimization problem, where the relationships between the material flows, compositions, operating procedures, event timing and various constraints are captured, and the optimization objectives are quantified. The problem is expressed in a standard mathematical form that is amenable to solution using commercial optimization software. The project explores optimization opportunities in order to identify optimal scheduling configurations that are not realizable based on human intuition due to the complexity inherent in the process.

Rack aware scheduling in HPC data centers: an energy conservation strategy

Energy consumption in high performance computing data centers has become a long standing issue. With rising costs of operating the data center, various techniques need to be employed to reduce the overall energy consumption. Currently, among others there are techniques that guarantee reduced energy consumption by powering on/off the idle nodes. However, most of them do not consider the energy consumed by other components in a rack. Our study addresses this aspect of the data center. We show that we can gain considerable energy savings by reducing the energy consumed by these rack components. In this regard, we propose a scheduling technique that will help schedule jobs with the above mentioned goal. We claim that by our scheduling technique we can reduce the energy consumption considerably without affecting other performance metrics of a job. We implement this technique as an enhancement to the well-known Maui scheduler and present our results. We propose three different algorithms as part of this technique. The algorithms evaluate the various trade-offs that could be possibly made with respect to overall cluster performance. We compare our technique with various currently available Maui scheduler configurations. We simulate a wide variety of workloads from real cluster deployments using the simulation mode of Maui. Our results consistently show about 7 to 14 % savings over the currently available Maui scheduler configurations. We shall also see that our technique can be applied in tandem with most of the existing energy aware scheduling techniques to achieve enhanced energy savings. We also consider the side effects of power losses due to the network switches as a result of deploying our technique. We compare our technique with the existing techniques in terms of the power losses due to these switches based on the results in Sharma and Ranganathan, Lecture Notes in Computer Science, vol. 5550, 2009 and account for the power losses. We there on provide a best fit scheme with the rack considerations. We then propose an enhanced technique that merges the two extremes of node allocation based on rack information. We see that we can provide a way to configure the scheduler based on the kind of workload that it schedules and reduce the effect of job splitting across multiple racks. We further discuss how the enhancement can be utilized to build a learning model which can be used to adaptively adjust the scheduling parameters based on the workload experienced.

A constraint programming-based solution approach for medical resident scheduling problems

Persistent calls come from within the graduate medical education community and from external sources for regulating the resident duty hours in order to meet the obligations about the quality of resident education, the well-being of residents themselves, and the quality of patient care services. The report of the Accreditation Council for Graduate Medical Education (ACGME) proposes common program requirements for resident hours. In this paper, we first develop a mixed-integer programming model for scheduling residents’ duty hours considering the on-call night, day-off, rest period, and total work-hour ACGME regulations as well as the demand coverage requirements of the residency program. Subsequently, we propose a column generation model that consists of a master problem and an auxiliary problem. The master problem finds a configuration of individual schedules that minimizes the sum of deviations from the desired service levels for the day and night periods. The formulation of this problem is possible by representing the feasible schedules using column variables, whereas the auxiliary problem finds the whole set of feasible schedules using constraint programming. The proposed approach has been tested on a series of problems using real data obtained from a hospital. The results indicate that high-quality schedules can be obtained within a few seconds.

Empirical performance evaluation of schedulers for cluster of workstations

Cluster computing is receiving exponential popularity as a choice for high performance computing. This is mainly due to its effective cost performance ratio. Resource management systems (RMS) are the key component to manage the resources of clusters efficiently and have a very vital role in the performance of distributed parallel systems especially a job scheduling module. In this paper, we have empirically evaluated four resource management systems (SGE, TORQUE, and MAUI Scheduler and SLURM) with special focus on job scheduler component. These schedulers have been evaluated on a more comprehensive set of metrics such as throughput, CPU, memory and network utilization. Experiments were carried out on three different size testbeds with a range of scheduler configurations such as FCFS, Backfilling, Fair share and SJF scheduling techniques. A head-to-head comparison of different scheduling techniques has also been presented which highlights the effect of RMS on the performance of scheduling techniques. It has been observed from results that relative difference among the performance of scheduling techniques reached up to 63%. We conclude from the experiments that there is no single choice of RMS which can be identified as the best but SLURM performs better than others in most of the cases.

Capacity and Outage Rate of OFDMA Cellular System with Fractional Frequency Reuse

Employing fractional frequency reuse (FFR) in OFDMA cellular systems is very attractive since it offers large capacity and single cell frequency reuse. However, its performance in practical environments, e.g. scheduling and arbitrary cell configurations, has not been well revealed. This paper analyzes the theoretical capacity and outage rate of an OFDMA cellular system employing FFR. Numerical examples show that FFR achieves higher capacity than the non-FFR equivalent when the outage rate is low.

Performance evaluation comparison of Snort NIDS under Linux and Windows Server

In this paper, we present an experimental evaluation and comparison of the performance of Snort NIDS when running under the two popular platforms of Linux and Windows 2003 Server. Snort's performance is measured when subjecting a PC host running Snort to both normal and malicious traffic, and with different traffic load conditions. Snort's performance is evaluated and compared in terms of throughput and packet loss. In order to offer sound interpretations and get better insight into the behavior of Snort, we also measure the packet loss encountered at the kernel level. In addition, we identify key system parameters (for both Linux and Windows) that provide a fine-grained control over the percentage of the CPU bandwidth allocated to Snort application and can consequently impact its performance. We investigate such an impact, and determine the most appropriate values to improve and optimize Snort's performance. Specifically, for Windows we investigate the impact of customizing the Processor Scheduling configuration option; and for Linux, we investigate the impact of tuning the Budget configurable parameter used in the Linux kernel's packet reception mechanism.

Modeling Framework for Airline Competition in the U.S. Domestic Network

An equilibrium-based modeling framework for airline competition analysis at the network level was developed. The framework considers a deregulated business environment in which air carriers can freely plan their networks and flight schedules to maximize their profits. Air carriers are assumed to reach a state of equilibrium, in which no air carrier can improve its profit by unilaterally modifying its network structure or schedule configuration. An equivalent mathematical program and an efficient solution algorithm are presented. The output of this program is the air carriers’ flight frequency for each market at equilibrium. The model was applied to the U.S. domestic network considering 13 major air carriers, which are assumed to compete in the markets between 236 metropolitan areas. The framework can be applied to answer a wide range of policy questions and to understand decisions related to airport congestion pricing, market access restrictions, air carriers merging, and so forth.

Metaheuristic scheduling of multiple picking agents for warehouse management

PurposeThis paper aims to develop a scheduler for multiple picking agents in a warehouse that takes into account distance and loading queue delay minimization within the context of minimizing makespan (i.e. picking time).Design/methodology/approachThe paper uses tabu search to solve the scheduling problem in a more global sense. Each search iteration is enhanced by a custom local search (LS) procedure that hastens convergence by driving a given schedule configuration quickly to a local minimum. In particular, basic operators transfer demand among agents to balance load and minimize makespan. The new load distribution is further improved by considering a vehicle‐routing problem on the picking assignments of the agents with relocated demands. Loading queue delays that may arise from the reassignments are systematically minimized using a fast scheduling heuristic.FindingsThe proposed tabu scheduler greatly improves over a widely practiced scheduling procedure for the given problem. Variants of the tabu scheduler produce solutions that are roughly of the same quality but exhibit considerable differences in computational time.Research limitations/implicationsThe proposed methodology is applicable only to the static scheduling problem where all inputs are known beforehand. Furthermore, of the possible delays during picking, only loading queues are explicitly addressed (although this is justifiable, given that these delays are dominant in the problem).Practical implicationsThe proposed approach can significantly increase through‐put and productivity in picking systems that utilize multiple intelligent agents (human pickers included), e.g. in warehouses/distribution centers.Originality/valueThe paper addresses a practical scheduling problem with a high degree of complexity, i.e. scheduler explicitly deals with delays while trying to minimize makespan (generally, delays are ignored in the literature to simplify things). In the tabu implementation, an LS procedure is introduced in the metaheuristic loop that enhances the search process by minimizing non‐productive time of picking agents (travel time and delays).

On the Robustness of Scheduling Against Channel Variations

In this paper, we study the robustness of scheduling against channel variations. Specifically, we consider a class of generalized proportional fairness (PF) schedulers that achieve optimal throughput-fairness trade-off and study their robustness in terms of maintaining a certain fairness criterion when the channel statistics change from the additive white Gaussian noise (AWGN) to Rayleigh fading for all users. We conclude that the scheduler is robust only when it is configured to achieve the proportional fairness. For the other scheduler configurations, which are not robust, we show how fairness deviates from the ideal one when the channel statistics change and show how scheduler parameters need to be adjusted to restore the desired fairness. We also show simulation results.

Router Scheduling Configuration Based on the Maximization of Benefit and Carried Best Effort Traffic

This paper shows a configuration scheme for networks with WFQ schedulers. It guarantees maximum revenue for the service provider in the worst case of network congestion. We focus on best effort traffic and select those flows that maximize the benefit while keeping the network utilization high. We show that optimum network configuration is feasible based only on knowledge of the topology. Its dependence on the pricing scheme can be reduced and even eliminated. We offer a formulation that reaches a tradeoff between network utilization, fairness, and user satisfaction.