Multi-class Queuing Network Model Research Articles

User learning can affect software system's response time

A session-based software system provides various services to its end users through graphical user interfaces. A novice user of a service's user interface takes more think time—the average time to comprehend the content and its layout on the interface—in comparison to expert users. The think time gradually decreases as she repeatedly comprehends the same interface over time. This decrease in think time is the user learning phenomenon. Owing to this learning behavior, the proportion of users—at various learning levels for different services—changes dynamically leading to a difference in the workload. Traditionally though, workload specifications (required for system performance evaluation) never accounted for user learning behavior. They generally assumed a global mean think time, instead. In this work, we first report an experimental study that demonstrates the impact of user learning of the graphical user interfaces of session-based systems. We measure the performance data of a real system while it is getting used by synthetic users (who are) at various learning levels. Statistical tests on our measured data reveal that different human learning curves, combined with different amounts of users arriving the system with various learning levels for different services, produce different system performance. Next, we explain a queueing network (QN) performance model called CogQN that accounts for user learning. It is a multi-class QN model where each service and its learning level constitute a class of users for the service. We present the concept of learning-level dependent, class-switching probabilities to model the probabilistic transition of learning levels of users for different services. This novel concept enables us to avoid the exponential explosion of classes that would otherwise be suffered by traditional class-switching theory, had it been used in this work. We solve the model using discrete-event simulation and validate it against the empirical data. Our CogQN model captures the effect of user learning on overall mean response times across different learning conditions within 10% absolute error in comparison to empirical data.

OptiSpot: minimizing application deployment cost using spot cloud resources.

The spot instance model is a virtual machine pricing scheme in which some resources of cloud providers are offered to the highest bidder. This leads to the formation of a spot price, whose fluctuations can determine customers to be overbid by other users and lose the virtual machine they rented. In this paper we propose OptiSpot, a heuristic to automate application deployment decisions on cloud providers that offer the spot pricing model. In particular, with our approach it is possible to determine: (i) which and how many resources to rent in order to run a cloud application, (ii) how to map the application components to the rented resources, and (iii) what spot price bids to use to minimize the total cost while maintaining an acceptable level of performance. To drive the decision making, our algorithm combines a multi-class queueing network model of the application with a Markov model that describes the stochastic evolution of the spot price and its influence on virtual machine reliability. We show, using a model developed for a real enterprise application and historical traces of the Amazon EC2 spot instance prices, that our heuristic finds low cost solutions that indeed guarantee the required levels of performance. The performance of our heuristic method is compared to that of nonlinear programming and shown to markedly accelerate the finding of low-cost optimal solutions.

Stability of multi-class queueing networks with infinite virtual queues

We generalize the standard multi-class queueing network model by allowing both standard queues and infinite virtual queues which have an infinite supply of work. We pose the general problem of finding policies which allow some of the nodes of the network to work with full utilization, and yet keep all the standard queues in the system stable. Toward this end we show that re-entrant lines, systems of two re-entrant lines through two service stations, and rings of service stations can be stabilized with priority policies under certain parameter restrictions. The analysis throughout the paper depends on model and policy and illustrates the difficulty in solving the general problem.

Routing Internet traffic in heterogeneous mesh networks: Analysis and algorithms

In practical wireless mesh networks (WMNs), gateways are subject to hard capacity limits on the aggregate number of flows (in terms of bit rate) that they can support. Thus, if traffic is routed in the mesh network without considering those constraints, as well as the traffic distribution, some gateways or intermediate mesh routers may rapidly get overloaded, and the network resources can be unevenly utilized. To address this problem, in this paper we firstly develop a multi-class queuing network model to analyze feasible throughput allocations, as well as average end-to-end delay, in heterogeneous WMNs. Guided by our analysis, we design a Capacity-Aware Route Selection algorithm ( CARS), which allocates network paths to downstream and upstream Internet flows so as to ensure a more balanced utilization of wireless network resources and gateways’ fixed connections. Through simulations in a number of different network scenarios we show that the CARS scheme significantly outperforms conventional shortest path routing, as well as an alternative routing method that distributes the traffic load on the gateway nodes to minimize its variance.

Performance analysis of multiclass queueing network model with positive Poisson batch sizes following discrete review policy

This paper is analyses one of the closed multiple classes queueing networks model by using the discrete review policies. The size of the batches are selected in the network according to discrete positive Poisson distribution and are served according to Exponential distribution. Our aim is to obtain model performance measures when the number of jobs in the system is fixed, the generating parameter of service times changes and also when the generating parameter of service times is fixed, The number ofjobs in the system changes. Finally, We calculate the efficiency measures of the network by using discrete-review policy results through an example.

Guest Editors' Introduction to the Special Issue on Quantitative Evaluation of Computer Systems

OFTWARE systems are expected to meet a multitude of quantitative constraints, such as timely response, performability, and efficient use of energy and other resources. Quantitative evaluation of systems refers to a set of techniques which can be applied to analyze such systems and obtain a range of quantitative characteristics of their performance, for example, the probability of satisfying a request within a given time or the appropriate number of replicas in a quorum air traffic control system. In this special issue, we publish a diverse range of novel research work which contributes to the development of quantitative evaluation of computer systems from a number of perspectives. These include theoretical foundations, models and modeling languages, analysis methods, simulation, verification, and software tools. Queueing networks have long been used for performance analysis of computer systems. This issue includes two papers related to queueing networks. In the first, “Enhanced Modeling and Solution of Layered Queueing Networks,” the layered queueing networks modeling formalism is presented. This modeling language allows a structured, hierarchical queueing network to be built up representing the resource usage and contention within a system. Analysis techniques which make use of the hierarchical structure allow complex models to be analyzed to address questions about the performance and dependability of the system. The second queueing paper, “CoMoM: Efficient Class-Oriented Evaluation of Multiclass Performance Models,” describes an improved solution technique for multiclass queueing network models. For more than 25 years, mean value analysis (MVA) has been one of the most frequently applied analysis methods of these models. This paper proposes an efficient rearrangement of recursive computational steps (referred to as the CoMoM method) such that it minimizes the number of recursive steps needed to solve the model. The CoMoM method makes possible the approximate analysis of much larger queueing network models, with tens of classes and thousands of nodes. Another well-established quantitative modeling technique is based on timed and stochastic extensions of Petri nets. In “State-Density Functions over DBM Domains in the Analysis of Non-Markovian Models,” a solution technique for such a Petri net extension is considered. The analysis of stochastic Petri nets with exponentially distributed firing times is well-developed and widely applied in software systemsmodeling.Incontrast,thecaseofstochasticPetrinets

Queueing analysis of a server node in transaction processing middleware systems

Quantitative performance modeling of complex information systems is of immense importance for designing enterprise e-business infrastructures and applications. In this paper, we present a traffic model of a server node in a typical transaction processing middleware system as well as a quantitative framework to model and analyze its performance. A multi-class open queueing network model is presented in which multi-class jobs are admitted to a number of server processes sharing hardware resources including the CPU and the disk. We have developed a viable approximation method, which decomposes the dependent components into their independent counterparts while preserving their relevant characteristics. We have conducted queueing-theoretic delay analyses and verified the approach using simulation. Results demonstrate the strength of our approach in predicting delays, elapsed times and other system performance measures.

Iterative algorithms for performance evaluation of closed network models

A number of nonlinear programming algorithms are proposed to obtain the approximate solutions for non-product form multiclass queueing network models, as well as priority queueing networks. Using sensitivity analysis, we develop an efficient iterative technique for closed queueing networks. We compare the approximate solutions obtained from our approach with the global balance solution. Examples illustrate the accuracy of the approximation, and compare the efficiency of the different optimization methods we have implemented.

Performance models of a proxy cache server: The impact of multimedia traffic

We propose a multi-class queuing network model to study the impact of multimedia traffic on the overall response times to Web requests for cases with and without a proxy cache server (PCS). We find that an increase of multimedia traffic percentage will significantly impact the response times for both cases with and without a PCS. The PCS “hit rate” is identified as one of the factors that impact the performance of a PCS for multimedia traffic. Further, we find that it may not pay to choose an overly-powerful PCS, as the marginal reduction in the overall response time cannot be justified. We show that the required time to travel through the client network represents the dominant factor of overall response time. Although the firm’s network bandwidth does not impact the decision of whether to install a PCS, improving it has a significant benefit of shortening the overall response time to users’ Web requests.

Modeling short-lived TCP connections with open multiclass queuing networks

In this paper we develop an open multiclass queuing network model to describe the behavior of short-lived TCP connections sharing a common IP network. The queuing network model is paired with a simple model of the IP network, and the two models are solved through an iterative procedure. The combined models need as inputs only the primitive network parameters, and they produce estimates of the packet loss probability, the round trip time, the TCP connection throughput, and of the average TCP connection completion time (that is, of the average time necessary to transfer a file with given size over a TCP connection). We derive models for both TCP-Tahoe and TCP-NewReno. The Tahoe model is presented in detail, while the NewReno model is presented describing differences with respect to Tahoe. Results are shown for both models. The analytical performance predictions are validated against detailed simulation experiments in realistic networking scenarios, proving that the proposed modeling approach is accurate.

Queueing network modelling and lead time compression of pharmaceutical drug development

Designing an optimized pharmaceutical drug development process is an important problem in itself and is of significant practical and research interest. Drug development lead time is a critical performance metric for a pharmaceutical company. In this paper, we develop a multiclass queueing network model to capture the project dynamics in drug development organizations that involve multiple, concurrent projects with contention for human/technical resources. We explore how drug development lead times can be reduced using efficient scheduling and critical mass-based resource management. The model captures important facets of any typical drug development organization, such as concurrent execution of multiple projects, contention for resources, feedback and reworking of project tasks, variability of new project initiations and task execution times, and certain scheduling issues. First, we show, using a class of fluctuation smoothing scheduling policies, that development lead times can be compressed impressively, without having to commit additional resources. Next, we show that critical mass-based project teams can compress lead times further. The model presented, though stylized, is sufficiently generic and conceptual, and will be of much value in new drug development project planning and management.

Lead time modeling and acceleration of product design and development

New product development is an important business process and constitutes a major contributor to the business excellence of any manufacturing firm. Designing an optimized new product development process is an important problem in itself and is of significant practical and research interest. Lead time is an important performance metric for a product development organization. We develop lead time models for product development organizations that involve multiple, concurrent projects with contention for human/technical resources. The objective is to explore how the lead times call be reduced using efficient scheduling, input control, load balancing, and variability reduction. The models are based on single class and multiclass queueing networks and capture important facets of a product development organization. Within the new product development process, we focus attention on the product design process. Two product design organizations, which we call Company ABC and Company XYZ, provide the real-world setting for our model-based lead time reduction. First, we present a coarse, conceptual queueing network model of Company ABC and show how rapid performance analysis can be used to explore opportunities for accelerating the design process. In particular, we show how effective input control, process control, load balancing, and cross-functional work can cut the lead times. Next, we present multiclass queueing network models for both the companies ABC and XYZ. The re-entrant line models show up certain scheduling issues pertaining to internal flows in the product design network. Using a class of fluctuation smoothing scheduling policies, we demonstrate how lead times can be reduced appreciably, without committing additional resources. The models presented are sufficiently generic and conceptual, and will be of much value in project planning and management in product design organizations and also more generally in product development organizations.

Approximate MVA solutions with fixed throughput classes

Transaction (open) type workloads are often used in approximating computer system workloads which are actually closed because open workloads provide reasonable estimates in many cases and their solutions are straight-forward. We have found that their use can distort the results for many workloads in a multiclass queueing network model of a computer system. We have replaced transaction workloads with what we call fixed class workloads. We present an approximate algorithm based on MVA that represents a class with a given throughput by a corresponding terminal or batch class, which we call a fixed class workload. We solve for the closed population required to deliver the requested throughput. We also present techniques for overcoming problems encountered in the solution of some fixed class models.

Scheduling semiconductor wafer fabrication

The impact that scheduling can have on the performance of semi-conductor wafer fabrication facilities is assessed. The performance measure considered is the mean throughput time (sometimes called cycle time, turnaround time or manufacturing interval) for a lot of wafers. A variety of input control and sequencing rules are evaluated using a simulation model of a representative, but fictitious, semiconductor wafer fabrication. Certain of these scheduling rules are derived by restricting attention to the sub-set of stations that are heavily utilized, and by using a Brownian network model, which approximates a multi-class queuing network model with dynamic control capability. Three versions of the wafer fabrication model, which differ only by the number of servers present at particular stations, are studied. The three versions have one, two, and four stations, respectively, that are heavily utilized (near 90% utilization). The simulation results indicate that scheduling has a significant impact on average throughput time, with larger improvements coming from discretionary imput control than from lot sequencing. The effects that specific sequencing rules have are highly dependent on both the type of input control used and the number of bottleneck stations in the fabrication.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Multiclass queueing networks with population constrainted subnetworks

A Multiclass Queueing Network model (MQN) is partitioned into a set of disjoint subnetworks. Population constraints are applied to each subnetwork such that within each subnetwork each population chain is either subject to an individual population constraint, or a group of chains may be subject to a common (shared) population constraint. Such population constraints are necessary in order to model multiprogramming level constraints in mainframe computer systems and window flow control mechanisms in computer communication networks. A computationally efficient approximate solution method is developed for solving MQN's with population constraints. Each subnetwork is reduced to a single approximately flow equivalent composite centre by assuming that the effect of other chains on a given chain can be adequately represented by their average customer populations. The accuracy of the population constraint approximation is compared against previous techniques by applying it to a set of test cases for which simulation solutions have previously been reported. The accuracy of the approximation technique is found to be good and in general is an improvement over previously published concurrency constraint approximations.