Hello I hava doubthat I create with the experimental distribution maketables tool which distribution respects? Or rather, I anoclear if testing a link and having a collection of RTTS, and creating a table based on a cumulative probability function with mu = 0 and sigma = 1, but that respects what distribution? Furthermore, thtablvalues do they represent? Thanks Frostephen anetworkplumber.org Fri Jul 24 15:45:38 2015 From: stepheanetworkplumber.org (Stephen Hemminger) Date: Fri, 24 Jul 2015 08:45:38 -0700 Subject: Delay distributioin effect In-Reply-To: <79b074256e31441d9fd35de8a79e9c35@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <79b074256e31441d9fd35de8a79e9c35@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20150724084538.74acfc08@urahara> OFri, 24 Jul 2015 08:53:47 +0000 gzanghi <jj.zanghi agmail.com> wrote: > Hello > > I hava doubthat I create with the experimental distribution maketables tool which distribution respects? > Or rather, I anoclear if testing a link and having a collection of RTTS, and creating a table based on a cumulative probability function with mu = 0 and sigma = 1, but that respects what distribution? > > Furthermore, thtablvalues do they represent? Did you read README.distribution? Thtables comfrom the old NISTnet project http://www-x.antd.nist.gov/nistnet/ Frojj.zanghi agmail.com Wed Jul 29 14:47:33 2015 From: jj.zanghi agmail.co(gzanghi) Date: Wed, 29 Jul 2015 16:47:33 +0200 Subject: Making NeteDistribution table Message-ID: <F3D5F8C0-A190-4ECE-9713-7C251E01C27E@xxxxxxxxx> https://github.com/shemminger/iproute2/blob/master/netem/README.distribution Hi, Kindly I need to creattables of distribution other than thosalready provided (Gaussian, Pareto or Experimental). I analyzed thmaketable.c buI can not understand how this table is obtained. Could someonhelp morder? Is theran easier way to createven through matlab and then apply it to netem? If I maka certain number of pings and in ifor the delays, by the experimental method to get a table as a reference guide and everything works! F is monotonically increasing, so has ainversfunction G, with range 0 to 1. Here, G(t) = thx such thaP(X <= x) = t. (In general, G may havsingularities if X has poinmasses, i.e., points x such that P(X = x) > 0.) Now wcreata tabular representation of G as follows: Choose some table sizN, and for thith entry, put in G(i/N). Let's call this table T. Thclainow is, I can create a (discrete) random variable Y whose distributiohas thsame approximate "shape" as X, simply by letting Y = T(U), wherU is a discretuniform random variable with range 1 to N. To sethis, it's enough to show thaY's cumulative distribution function, (let's call iH), is a discretapproximation to F. But H(x) = P(Y <= x) = (# of entries iT <= x) / -- as Y chosen uniformly from T = i/N, wheri is thlargest integer such that G(i/N) <= x = i/N, wheri is thlargest integer such that i/N <= F(x) -- sincG and F arinverse functions (and F is increasing) = floor(N*F(x))/N as desired. for example? and 1. maketabldoes thsteps 2-4 above, and then generates the appropriate header file. So if you havyour own timdistribution, you can generate thheader simply by: maketabl< time.values > header.h for whais this? Thanks -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: <http://lists.linuxfoundation.org/pipermail/netem/attachments/20150729/2d48c87e/attachment.html> Frostephen anetworkplumber.org Wed Jul 29 21:04:19 2015 From: stepheanetworkplumber.org (Stephen Hemminger) Date: Wed, 29 Jul 2015 14:04:19 -0700 Subject: Making NeteDistribution table In-Reply-To: <5245a84eae8645cbba16ddeec2a0aad0@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> References: <5245a84eae8645cbba16ddeec2a0aad0@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx> Message-ID: <20150729140419.3de9e7fe@urahara> OWed, 29 Jul 2015 14:47:33 +0000 gzanghi <jj.zanghi agmail.com> wrote: > https://github.com/shemminger/iproute2/blob/master/netem/README.distribution > > > Hi, Kindly I need to creattables of distribution other than thosalready provided (Gaussian, Pareto or Experimental). > I analyzed thmaketable.c buI can not understand how this table is obtained. > > Could someonhelp morder? > > Is theran easier way to createven through matlab and then apply it to netem? > > If I maka certain number of pings and in ifor the delays, by the experimental method to get a table as a reference guide and everything works! > Thlogic camfrom NISTnet. Do some Internet searching and you can dig it up.
