Michael Muskulus

Analysis and modeling of job arrivals in a production grid

In this paper we present an initial analysis of job arrivals in a production data-intensive Grid and investigate several traffic models for the interarrival time processes. Our analysis focuses on the heavy-tail behavior and autocorrelations, and the modeling is carried out at three different levels: Grid, Virtual Organization (VO), and region. A set of m-state Markov modulated Poisson processes (MMPP) is investigated, while Poisson processes and hyperexponential renewal processes are evaluated for comparison studies. We apply the transportation distance metric from dynamical systems theory to further characterize the differences between the data trace and the simulated time series, and estimate errors by bootstrapping. The experimental results show that MMPPs with a certain number of states are successful to a certain extent in simulating the job traffic at different levels, fitting both the interarrival time distribution and the autocorrelation function. However, MMPPs are not able to match the autocorrelations for certain VOs, in which strong deterministic semi-periodic patterns are observed. These patterns are further characterized using different representations. Future work is needed to model both deterministic and stochastic components in order to better capture the correlation structure in the series.

Fluctuations and determinism of respiratory impedance in asthma and chronic obstructive pulmonary disease

Asthma and COPD are chronic respiratory diseases that fluctuate widely with regard to clinical symptoms and airway obstruction, complicating treatment and prediction of exacerbations. Time series of respiratory impedance obtained by the forced oscillation technique are a convenient tool to study the respiratory system with high temporal resolution. In previous studies it was suggested that power-law-like fluctuations exist also in the healthy lung and that respiratory system impedance variability differs in asthma. In this study we elucidate such differences in a population of well-characterized subjects with asthma (n = 13, GINA 1+2), COPD (n = 12, GOLD I+II), and controls (n = 10) from time series at single frequency (12 min, f = 8 Hz). Maximum likelihood estimation did not rule out power-law behavior, accepting the null hypothesis in 17/35 cases (P > 0.05) and with significant differences in exponents for COPD (P < 0.03). Detrended fluctuation analysis exhibited scaling exponents close to 0.5, indicating few correlations, with no differences between groups (P > 0.14). In a second approach, we considered asthma and COPD as dynamic diseases, corresponding to changes of unknown parameters in a deterministic system. The similarity in shape between the combined probability distributions of normalized resistance and reactance was quantified by Wasserstein distances and reliably distinguished the two diseases (cross-validated predictive accuracy 0.80; sensitivity 0.83, specificity 0.77 for COPD). Wasserstein distances between 3+3 dimensional phase space reconstructions resulted in marginally better classification (accuracy 0.84, sensitivity 0.83, specificity 0.85). These latter findings suggest that the dynamics of respiratory impedance contain valuable information for the diagnosis and monitoring of patients with asthma and COPD, whereas the value of the stochastic approach is not clear presently.

Modeling job arrivals in a data-intensive grid

In this paper we present an initial analysis of job arrivals in a production data-intensive Grid and investigate several traffic models to characterize the interarrival time processes. Our analysis focuses on the heavy-tail behavior and autocorrelation structures, and the modeling is carried out at three different levels: Grid, Virtual Organization (VO), and region. A set of m-state Markov modulated Poisson processes (MMPP) is investigated, while Poisson processes and hyperexponential renewal processes are evaluated for comparison studies. We apply the transportation distance metric from dynamical systems theory to further characterize the differences between the data trace and the simulated time series, and estimate errors by bootstrapping. The experimental results show that MMPPs with a certain number of states are successful to a certain extent in simulating the job traffic at different levels, fitting both the interarrival time distribution and the autocorrelation function. However, MMPPs are not able to match the autocorrelations for certain VOs, in which strong deterministic semi-periodic patterns are observed. These patterns are further characterized using different representations. Future work is needed to model both deterministic and stochastic components in order to better capture the correlation structure in the series.

Modeling correlated workloads by combining model based clustering and a localized sampling algorithm

We propose a new model for workload attributes on spaceshared computer systems, which is able to fit both marginal distributions and second order statistics such as the autocorrelation function (ACF). The modeling process is formed by a two-stage approach: Firstly, a mixture of Gaussians model is used to fit the probability density function (PDF), whose parameters are estimated via a framework called model based clustering (MBC). The MBC framework can further cluster the data according to the Gaussian components, which plays an important role in creating correlations in the next stage. Secondly, a novel localized sampling algorithm is proposed to generate correlations in the synthetic data series. It is discovered that the number of repetitions of cluster labels obtained via MBC empirically follow a Zipf-like (power law) distribution. Sampling repeatedly from a certain cluster according to the Zipf law is able to create correlations in the series. Furthermore, a cluster permutation procedure is introduced so that the autocorrelations in the synthetic data can be controlled to match those in the real trace via a single parameter. Our approach can generalize to more than one dimension, which means multiple correlated workload attributes can be modeled simultaneously. Experimental studies are conducted to evaluate the proposed algorithm using real workload traces on production systems such as Grids and supercomputers.

Analysis and Synthesis of Pseudo-Periodic Job Arrivals in Grids: A Matching Pursuit Approach

Pseudo-periodicity is one of the basic job arrival patterns on data-intensive clusters and Grids. In this paper, a signal decomposition methodology called matching pursuit is applied for analysis and synthesis of pseudo-periodic job arrival processes. The matching pursuit decomposition is well localized both in time and frequency, and it is naturally suited for analyzing non-stationary as well as stationary signals. The stationarity of the processes can be quantitatively measured by permutation entropy, with which the relationship between stationarity and modeling complexity is excellently explained. Quantitative methods based on the power spectrum are also provided to measure the degree of periodicity present in the data. Matching pursuit is further shown to be able to extract patterns from signals, which is an attractive feature from a modeling perspective. Real world workload data from production clusters and Grids are used to empirically evaluate the proposed measures and methodologies.

Functional similarities and distance properties

The analysis of functional and effective brain connectivity forms an important tool for unraveling structure-function relationships from neurophysiological data. It has clinical applications, supports the formulation of hypotheses regarding the role and localization of functional processes, and is often an initial step in modeling. However, only a few of the commonly applied connectivity measures respect metric properties: reflexivity, symmetry, and the triangle inequality. This may hamper interpretation of findings and subsequent analysis. Connectivity indices obtained by metric measures can be seen as functional distances, and may be represented in Euclidean space by the methods of multidimensional scaling. We sketch some classes of measures that do allow for such a reconstruction, in particular the class of Wasserstein distances, and discuss their merits for interpreting cortical activity assessed by magnetoencephalography. In an application to magnetoencephalographic recordings during the execution of a bimanual task, the Wasserstein distances between relative circular variances indicated cortico-muscular synchrony as well as cross-talk between bilateral primary motor areas in the beta-band.

Strategies of loop recombination in ciliates

The concept of breakpoint graph, known from the theory of sorting by reversal, has been successfully applied in the theory of gene assembly in ciliates. We further investigate its usage for gene assembly, and show that the graph allows for an efficient characterization of the possible orders of loop recombination operations (one of the three types of molecular operations that accomplish gene assembly) for a given gene during gene assembly. The characterization is based on spanning trees within a graph built upon the connected components in the breakpoint graph. We work in the abstract and more general setting of so-called legal strings.

Fatigue reassessment for lifetime extension of offshore wind monopile substructures

Fatigue reassessment is required to decide about lifetime extension of aging offshore wind farms. This paper presents a methodology to identify important parameters to monitor during the operational phase of offshore wind turbines. An elementary effects method is applied to analyze the global sensitivity of residual fatigue lifetimes to environmental, structural and operational parameters. Therefore, renewed lifetime simulations are performed for a case study which consists of a 5 MW turbine with monopile substructure in 20 m water depth. Results show that corrosion, turbine availability, and turbulence intensity are the most influential parameters. This can vary strongly for other settings (water depth, turbine size, etc.) making case-specific assessments necessary.

The simultaneous effect of a fairing tower and increased blade flexibility on a downwind mounted rotor

This is a parametric study on how blade and tower loads for a prototypical downwind offshore wind turbine are affected as the tower geometry and blade properties are changed. Downwind turbines have the potential to reduce the cost of energy, as blades can be more flexible and lighter, but the tower shadow induces additional structural vibrations. In order to reduce the latter, a fairing around the tower has been introduced. The length of the fairing is varied, adjusting the rotor overhang accordingly. Additionally, the blade weight and stiffness are adjusted. The blade and tower fatigue loads are, thereby, significantly decreased. In the first case, a maximum reduction of 8% and 28% (for the blade root bending and tower bottom moment, respectively) was achieved, compared to a downwind version of the National Renewable Energy Laboratory (NREL) 5 MW reference wind turbine on a monopile tower. Using softer and lighter blades resulted in loads even lower than for the conventional upwind rotor of the NREL turb...

Cycles and communicating classes in membrane systems and molecular dynamics

Abstract We are considering sequential membrane systems and molecular dynamics from the viewpoint of Markov chain theory. The configuration space of these systems (including the transitions) is a special kind of directed graph, called a pseudo-lattice digraph, which is closely related to the stoichiometric matrix. Taking advantage of the monoidal structure of this space, we introduce the algebraic notion of precycle. A precycle leads to the identification of cycles by means of the concept of defect, which is a set of geometric constraints on configuration space. Two efficient algorithms for evaluating precycles and defects are given: one is an algorithm due to Contejean and Devie, the other is a novel branch-and-bound tree search procedure. Cycles partition configuration space into equivalence classes, called the communicating classes. The structure of the communicating classes in the free regime–where all rules are enabled–is analyzed: testing for communication can be done efficiently. We show how to apply these ideas to a biological regulatory system.