Package org.networkcalculus.dnc.algebra.disco.con_pw_affine

Class Deconvolution_Disco_ConPwAffine

  • public abstract class Deconvolution_Disco_ConPwAffine
extends java.lang.Object
    Deconvolution for convex or (almost) concave, (almost) continuous functions. Functions of this class are defined by - the maximum over multiple rate-latency curves (mRL, convex functions), - the minimum over multiple token-bucket curves (mTB, almost concave/continuous functions as f(0)=0), - the delayed infinite burst functions \delta_{T} used in the min/max above. The deconvolution of piecewise affine curves implemented in the original dnclib was based on a flawed result (for pw affine but working for affine curves) presented in Improving Performance Bounds in Feed-Forward Networks by Paying Multiplexing Only Once (Jens Schmitt, Frank A. Zdarsky, Ivan Martinovic), In Proc. of the 14th GI/ITG MMB, 2008. This replacement, implemented since the version 2.2.5 (2015-Sep-29), is however restricted to the above class of curves.

    • Constructor Detail

      • Deconvolution_Disco_ConPwAffine

        public Deconvolution_Disco_ConPwAffine()

    • Method Detail

      • deconvolve_almostConcCs_SCs

        public static java.util.Set<ArrivalCurve> deconvolve_almostConcCs_SCs​(java.util.Set<Curve> curves,
                                                                      java.util.Set<ServiceCurve> service_curves)

      • deconvolve_con_pw_affine

        private static ArrivalCurve deconvolve_con_pw_affine​(Curve curve_1,
                                                     Curve curve_2)
        Returns the deconvolution of an (almost) concave arrival curve and a convex service curve.
        Parameters:
        curve_1 - The (almost) concave arrival curve.
        curve_2 - The convex service curve.
        Returns:
        The deconvolved curve, an arrival curve.