Giuliano Casale

Bottlenecks identification in multiclass queueing networks using convex polytopes

It is known that the resources that limit the overall performance of a system are the congested ones, referred to as bottlenecks. From knowledge of bottleneck stations, it is possible, with limited computational effort, to derive asymptotic values of several performance indices. While identifying the bottleneck stations is a well-established practice under a single-class workload, no simple methodology for multiclass models exists. We present new algorithms for identifying the bottlenecks in multiclass queueing networks with constant-rate servers. We show how the application of assessed techniques, such as the ones related to convex polytopes, can provide insights on the performance of a queueing network. The application of our techniques to the asymptotic analysis of closed product-form networks is also investigated.

The JMT Simulator for Performance Evaluation of Non-Product-Form Queueing Networks

This paper describes JSIM: the simulation module of the Java Modelling Tools (JMT), an open-source fully-portable Java suite for capacity planning studies. The simulator has been purposely developed to help both unexperienced and advanced users. Most of the difficult decisions that are needed in order to run simulations properly, such as the detection of the transient part of samples to be discarded, have been automated. The tool also provides guidance over the graphical design of the network and over the analysis and the plot of the results. What-if parametric analyses for parametric evaluation of complex systems are supported. Several features that increase the generality of the applications to capacity planning studies are provided, among them fork-join service centers, regions with finite capacity, state-dependent routing algorithms, priority classes and import of real workload distributions from log files

A two-phase relaxation-based heuristic for the maximum feasible subsystem problem

We consider the maximum feasible subsystem problem in which, given an infeasible system of linear inequalities, one wishes to determine a largest feasible subsystem. The focus is on the version with bounded variables that naturally arises in several fields of application. To tackle this NP-hard problem, we propose a simple but efficient two-phase relaxation-based heuristic. First a feasible subsystem is derived from a relaxation (linearization) of an exact continuous bilinear formulation, and then a smaller subproblem is solved to optimality in order to identify all other inequalities that can be added to the current feasible subsystem while preserving feasibility. Computational results, reported for several classes of instances, arising from classification and telecommunication applications, indicate that our method compares well with one of the best available heuristics and with state-of-the-art exact algorithms.

Optimal number of nodes for computation in grid environments

We show that there exists an optimal number of nodes to be assigned to jobs for execution in grid systems, which depends on the distributions of computation and communication service times. We also show that the analytical models proposed for parallel computers are not accurate for grid systems. We therefore extend to grid environment the definitions of speedup, efficiency and efficacy that are usually given for parallel systems. We also adopt a queueing network model with three different types of workload to prove that in every case an optimal number of nodes exists and that the mean value of CPU and communication service times is just a scale factor for this optimum.

Combining queueing networks and web usage mining techniques for web performance analysis

The growing interest in Web applications that satisfy end-to-end Quality of Service (QoS) requirements is leading many organizations to build and analyze performance behavior models. In this direction, Web usage mining techniques may help in the automatic construction of user profiles from Web access logs. However, their use has been mainly limited to customer relationship management (CRM) issues and to market analyses. The aim of this paper is to explain how Web usage mining can be combined with queueing networks for effective Web capacity planning. After introducing a new general relative cosine similarity measure, we define a performance-oriented similarity for Web usage data. A methodology to devise the input parameters of a queueing network from the resulting clusters is also presented. Finally, the proposed approach is illustrated on a simple case study.

Performance Issues in Video Streaming Environments

The main motivation for applying capacity planning techniques to the design of Internet-based video services is that very often these systems incur significant delays that decrease the offered Quality of Service below an acceptable level. However, the very high fluctuations of the Internet traffic make the identification of the correct capacity of the server and network components a very hard task to be accomplished. In this paper we analyze the metrics used to evaluate the performance perceived by users of video streaming applications and we introduce and validate some new metrics. These variables may be used to parameterize the models required by capacity planning studies. A characterization of users’ behavior for live and stored video streams is presented. The knowledge of the behavior of these two streams is important to identify the capacity to be allocated in order to satisfy the QoS requirements. Several issues concerning the security of video streaming are also discussed.

A New Class of Non-Iterative Bounds for Closed Queueing Networks

A new emerging class of problems related to the online configuration and optimization of computer systems and networks requires the solution in a very short amount of time of a large number of analytical performance models, often based on queueing networks. In this paper we propose the Geometric Bounds (GB), a new family of fast noniterative bounds on performance metrics of closed productform queueing networks. In spite of their simplicity, the proposed bounds are more accurate than the popular Balanced Job Bounds (BJB), even in the difficult case of networks with multiple bottlenecks or large delays.

GIVS: integrity validation for grid security

In this paper we address the problem of granting the correctness of Grid computations. We introduce a Grid Integrity Validation Scheme (GIVS) that may reveal the presence of malicious hosts by statistical sampling of computations results. Performance overheads of GIVS strategies are evaluated using statistical models and simulation.