Dariusz R Augustyn

A Choice of Optimal Packet Dropping Function for Active Queue Management

Algorithms of queue management in IP routers determine which packet should be deleted when necessary. The article investigates the influence of packet rejection probability function on the performance, i.e. response time for in case NLRED queues. For standard NLRED the linear packet dropping function is replaced by a quadratic function. Solution presented in this article based on orthogonal polynomials. Depending on a particular criterion of optimization a proper non-linear function may be chosen, but using averaged criterion leads up to choose the linear one as an optimal one.

Applying Advanced Methods of Query Selectivity Estimation in Oracle DBMS

The paper shows the solution of the query selectivity estimation problem for certain types of database queries with a selection condition based on several table attributes. The selectivity parameter allows for estimating a size of data satisfying a query condition. An estimator of a multidimensional probability density function is required for an accurate selectivity calculation for conditions involving many attributes and correlated attribute values. Using multidimensional histogram as a nonparametric density function estimator is mostly too much storage-consuming. The implementation of the known unconventional storage-efficient approach based on Discrete Cosine Transform spectrum of a multidimensional histogram is presented. This solution extends functionality of the Oracle DBMS cost-based query optimizer. The method of experimental obtaining error-optimal parameters values of spectrum storage for typical attributes value distributions is considered.

Applying CUDA Technology in DCT-Based Method of Query Selectivity Estimation

The problem of efficient calculation of query selectivity estimation is considered in this paper. The selectivity parameter allows database query optimizer to estimate the size of the data satisfying given condition, which is needed to choose the best query execution plan. Obtaining query selectivity in case of a multi-attribute selection condition requires a representation of multidimensional attributes values distribution. This paper describes in details solution of this problem, which utilizes Discrete Cosine Transform and CUDA-based algorithm for obtaining selectivity estimation. There are also some remarks about efficiency and advantages of this approach.

A RED modified weighted moving average for soft real-time application

Abstract The popularity of TCP/IP has resulted in an increase in usage of best-effort networks for real-time communication. Much effort has been spent to ensure quality of service for soft real-time traffic over IP networks. The Internet Engineering Task Force has proposed some architecture components, such as Active Queue Management (AQM). The paper investigates the influence of the weighted moving average on packet waiting time reduction for an AQM mechanism: the RED algorithm. The proposed method for computing the average queue length is based on a difference equation (a recursive equation). Depending on a particular optimality criterion, proper parameters of the modified weighted moving average function can be chosen. This change will allow reducing the number of violations of timing constraints and better use of this mechanism for soft real-time transmissions. The optimization problem is solved through simulations performed in OMNeT++ and later verified experimentally on a Linux implementation

The Impact of the Modified Weighted Moving Average on the Performance of the RED Mechanism

Algorithms of queue management in IP routers determine which packet should be deleted when necessary. The article investigates the influence of weighted moving average on the performance of the RED mechanism. The proposed average queue length calculating method is based on a difference equation (a recursive equation). Depending on a particular criterion of optimization a proper parameters of modified weighted moving average function may be chosen.

Cloud Service Solving N-Body Problem Based on Windows Azure Platform

This paper shows how to use cloud computing to solve N-body problem. It presents an idea and implementation of cloud service based on Windows Azure Platform. Clients can access cloud service via Internet over HTTP protocol. They create computation tasks supplying simulation parameters such as number of steps, time step and XML file with body definitions (initial position, mass and velocity). Presented solution uses Barnes-Hut Algorithm (based on adaptive oct tree) to reduce computation complexity form N ×N to N logN. All body interactions are computed in parallel, on worker nodes in cloud.

Query-Condition-Aware Histograms in Selectivity Estimation Method

The paper shows an adaptive approach to the query selectivity estimation problem for queries with a range selection condition based on continuous attributes. The selectivity factor estimates a size of data satisfying a query condition. This estimation is calculated at the initial stage of the query processing for choosing the optimal query execution plan. A non-parametric estimator of probability density of attribute values distribution is required for the selectivity calculation. Most of known approaches use equi-width or equi-height histograms as representations of attribute values distributions. The proposed approach uses a new type of histogram based on either an attribute values distribution or a distribution of range bounds of a query selection condition. Applying query-condition-aware histogram lets obtain more accurate selectivity values than using a standard histogram. The approach may be implemented as some extension of query optimizer of DBMS Oracle using ODCI Stats module.

Using the Model of Continuous Dynamical System with Viscous Resistance Forces for Improving Distribution Prediction Based on Evolution of Quantiles

The paper considers the problem of prediction of a probability distribution. We take into account an extrapolation model based on evolution of quantiles. We may use any concrete model which allows to track and extrapolate boundaries of buckets of an equi-height histogram. This histogram with p + 1 boundaries is equivalent to p-quantiles. Using such baseline extrapolation model we may obtain lines of locations of bucket boundaries that may intersect in future. To avoid intersections and to extend (in time) correctness of the results, we propose to use a model of continuous dynamical system with viscous resistance forces for obtaining improved lines of locations. The proposed model allows to obtain lines with unchanged shapes or very similar ones (comparing to the results from the baseline extrapolation model) but without any intersections. This approach will be helpful when a previously used baseline extrapolation model is too much time limited. The work was inspired by the problem of prediction of an attribute value distribution used for query selectivity estimation. However, the proposed method may be applied not only in query optimization problem.

GPU-Accelerated Method of Query Selectivity Estimation for Non Equi-Join Conditions Based on Discrete Fourier Transform

Selectivity factor is obtained by database query optimizer for estimating the size of data that satisfy a query condition. This allows to choose the optimal query execution plan. In this paper we consider the problem of selectivity estimation for inequality predicates based on two attributes, therefore the proposed solution allows to estimate the size of data that satisfy theta-join conditions. The proposed method is based on Discrete Fourier Transform and convolution theorem. DFT spectrums are used as representations of distribution of attribute values. We compute selectivity either performing Inverse DFT (for an inequality condition based on two attributes) or avoiding it (for a single-attribute range one). Selectivity calculation is a time-critical operation performed during an on-line query preparing phase. We show that by applying parallel processing capabilities of Graphical Processing Unit, the implementation of the method satisfies the assumed time constraint.

The Method of Query Selectivity Estimation for Selection Conditions Based on Sum of Sub-Independent Attributes

Selectivity estimation is an activity performed during a query optimization process. Selectivity parameter lets estimate the query result size before the query is really executed. This allows to obtain the best query execution plan. For complex queries (where selection condition is based on many attributes) an accurate selectivity estimation requires a multidimensional distribution of attributes values. But often, attribute value independence assumption and usage of only 1-dimensional distributions give a sufficient accuracy of selectivity approximation. The paper describes the method of selectivity estimation for queries with a complex selection condition based on a sum of independent attributes or sub-independent ones. The proposed method operates on 1-dimensional Fourier Transforms of marginal distributions of attributes that are involved in the selection condition.