X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/7ec790f5a5d5c9f82f7f984cbea4a5b886968902..c2c36bb9b8f9f004079e2eb3064ccd25191ada87:/src/surf/lagrange.c diff --git a/src/surf/lagrange.c b/src/surf/lagrange.c index 01e22449c2..34a0df7c7e 100644 --- a/src/surf/lagrange.c +++ b/src/surf/lagrange.c @@ -1,10 +1,7 @@ /* $Id$ */ - /* Copyright (c) 2007 Arnaud Legrand, Pedro Velho. All rights reserved. */ - /* This program is free software; you can redistribute it and/or modify it * under the terms of the license (GNU LGPL) which comes with this package. */ - /* * Modelling the proportional fairness using the Lagrange Optimization * Approach. For a detailed description see: @@ -20,69 +17,125 @@ #include #endif -#define LAMBDA_STEP 0.01 - XBT_LOG_NEW_DEFAULT_SUBCATEGORY(surf_lagrange, surf, "Logging specific to SURF (lagrange)"); +XBT_LOG_NEW_SUBCATEGORY(surf_lagrange_dichotomy, surf, + "Logging specific to SURF (lagrange dichotomy)"); -XBT_LOG_NEW_SUBCATEGORY(surf_writelambda, surf, - "Generates the lambda.in file. WARNING: the size of this file might be a few GBs."); - +/* + * Local prototypes to implement the lagrangian optimization with optimal step, also called dichotomy. + */ +//solves the proportional fairness using a lagrange optimizition with dichotomy step void lagrange_solve(lmm_system_t sys); +//computes the value of the dichotomy using a initial values, init, with a specific variable or constraint +double dichotomy(double init, double diff(double, void *), void *var_cnst, + double min_error); +//computes the value of the differential of variable param_var applied to mu +double partial_diff_mu(double mu, void *param_var); +//computes the value of the differential of constraint param_cnst applied to lambda +double partial_diff_lambda(double lambda, void *param_cnst); +//auxiliar function to compute the partial_diff +double diff_aux(lmm_variable_t var, double x); + + +static int __check_kkt(xbt_swag_t cnst_list, xbt_swag_t var_list, int warn) +{ + xbt_swag_t elem_list = NULL; + lmm_element_t elem = NULL; + lmm_constraint_t cnst = NULL; + lmm_variable_t var = NULL; + + double tmp; + + //verify the KKT property for each link + xbt_swag_foreach(cnst, cnst_list) { + tmp = 0; + elem_list = &(cnst->element_set); + xbt_swag_foreach(elem, elem_list) { + var = elem->variable; + if (var->weight <= 0) + continue; + tmp += var->value; + } + + if (double_positive(tmp - cnst->bound)) { + if (warn) + WARN3 + ("The link (%p) is over-used. Expected less than %f and got %f", + cnst, cnst->bound, tmp); + return 0; + } + DEBUG3("Checking KKT for constraint (%p): sat = %f, lambda = %f ", + cnst, tmp - cnst->bound, cnst->lambda); + +/* if(!((fabs(tmp - cnst->bound)lambda>=MAXMIN_PRECISION) || */ +/* (fabs(tmp - cnst->bound)>=MAXMIN_PRECISION && cnst->lambdabound < 0 || var->weight <= 0) + continue; + DEBUG3("Checking KKT for variable (%p): sat = %f mu = %f", var, + var->value - var->bound, var->mu); + + if (double_positive(var->value - var->bound)) { + if (warn) + WARN3 + ("The variable (%p) is too large. Expected less than %f and got %f", + var, var->bound, var->value); + return 0; + } + +/* if(!((fabs(var->value - var->bound)mu>=MAXMIN_PRECISION) || */ +/* (fabs(var->value - var->bound)>=MAXMIN_PRECISION && var->mu"); - DEBUG1("#### Maximum number of iterations : %d", max_iterations); - DEBUG1("#### Minimum error tolerated : %e", epsilon_min_error); - DEBUG1("#### Step : %e", sigma_step); - + DEBUG1("#### Maximum number of iterations : %d", max_iterations); + DEBUG1("#### Minimum error tolerated : %e", + epsilon_min_error); + DEBUG1("#### Minimum error tolerated (dichotomy) : %e", + dichotomy_min_error); - if ( !(sys->modified)) + if (!(sys->modified)) return; /* @@ -90,240 +143,415 @@ void lagrange_solve(lmm_system_t sys) * Associate an index in the swag variables. Initialize mu. */ var_list = &(sys->variable_set); - i=0; - xbt_swag_foreach(var1, var_list) { - if((var1->bound < 0.0) || (var1->weight <= 0.0)){ - DEBUG1("#### NOTE var1(%d) is a boundless variable", i); - var1->mu = -1.0; - } else{ - var1->mu = 1.0; - var1->new_mu = 2.0; + i = 0; + xbt_swag_foreach(var, var_list) { + if ((var->bound < 0.0) || (var->weight <= 0.0)) { + DEBUG1("#### NOTE var(%d) is a boundless (or inactive) variable", i); + var->mu = -1.0; + } else { + var->mu = 1.0; + var->new_mu = 2.0; } - DEBUG2("#### var1(%d)->mu: %e", i, var1->mu); - DEBUG2("#### var1(%d)->weight: %e", i, var1->weight); + DEBUG3("#### var(%d) %p ->mu : %e", i, var, var->mu); + DEBUG3("#### var(%d) %p ->weight: %e", i, var, var->weight); + DEBUG3("#### var(%d) %p ->bound: %e", i, var, var->bound); i++; } /* * Initialize lambda. */ - cnst_list=&(sys->active_constraint_set); - xbt_swag_foreach(cnst1, cnst_list) { - cnst1->lambda = 1.0; - cnst1->new_lambda = 2.0; - DEBUG2("#### cnst1(%p)->lambda: %e", cnst1, cnst1->lambda); - } - - if(XBT_LOG_ISENABLED(surf_writelambda, xbt_log_priority_debug)) { - gnuplot_file = fopen("lambda.in", "w"); - fprintf(gnuplot_file, "# iteration lambda1 lambda2 lambda3 ... lambdaP\n"); + cnst_list = &(sys->active_constraint_set); + xbt_swag_foreach(cnst, cnst_list) { + cnst->lambda = 1.0; + cnst->new_lambda = 2.0; + DEBUG2("#### cnst(%p)->lambda : %e", cnst, cnst->lambda); } - /* * While doesn't reach a minimun error or a number maximum of iterations. */ - while(overall_error > epsilon_min_error && iteration < max_iterations){ + while (overall_error > epsilon_min_error && iteration < max_iterations) { + iteration++; + DEBUG1("************** ITERATION %d **************", iteration); - /* d Dual - * Compute the value of ----------- (\lambda^k, \mu^k) this portion - * d \mu_i^k - * of code depends on function f(x). + /* + * Compute the value of mu_i */ - var_list = &(sys->variable_set); - xbt_swag_foreach(var1, var_list) { - mu_partial = 0; - if((var1->bound >= 0) && (var1->weight > 0) ){ - //for each link with capacity cnsts[i] that uses flow of variable var1 do - for(i=0; icnsts_number; i++) - mu_partial += (var1->cnsts[i].constraint)->lambda + var1->mu; - - mu_partial = -1.0 / mu_partial + var1->bound; - var1->new_mu = var1->mu - sigma_step * mu_partial; - - if(var1->new_mu < 0){ - var1->new_mu = 0; - } + //forall mu_i in mu_1, mu_2, ..., mu_n + xbt_swag_foreach(var, var_list) { + if ((var->bound >= 0) && (var->weight > 0)) { + DEBUG1("====> Working on var (%p)", var); + var->new_mu = + dichotomy(var->mu, partial_diff_mu, var, dichotomy_min_error); + if (var->new_mu < 0) + var->new_mu = 0; + DEBUG3("====> var->mu (%p) : %g -> %g", var, var->mu, var->new_mu); + var->mu = var->new_mu; } } - - /* d Dual - * Compute the value of ------------- (\lambda^k, \mu^k) this portion - * d \lambda_i^k - * of code depends on function f(x). + /* + * Compute the value of lambda_i */ - j=0; - if(XBT_LOG_ISENABLED(surf_writelambda, xbt_log_priority_debug)) { - fprintf(gnuplot_file, "\n%d",iteration); + //forall lambda_i in lambda_1, lambda_2, ..., lambda_n + xbt_swag_foreach(cnst, cnst_list) { + DEBUG1("====> Working on cnst (%p)", cnst); + cnst->new_lambda = + dichotomy(cnst->lambda, partial_diff_lambda, cnst, + dichotomy_min_error); + DEBUG2("====> cnst->lambda (%p) = %e", cnst, cnst->new_lambda); + cnst->lambda = cnst->new_lambda; } - xbt_swag_foreach(cnst1, cnst_list) { - j++; - - lambda_partial = 0; - - elem_list = &(cnst1->element_set); - watch_out=0; - xbt_swag_foreach(elem1, elem_list) { - - var2 = elem1->variable; - - if(var2->weight<=0) continue; + /* + * Now computes the values of each variable (\rho) based on + * the values of \lambda and \mu. + */ + overall_error = 0; + xbt_swag_foreach(var, var_list) { + if (var->weight <= 0) + var->value = 0.0; + else { + //compute sigma_i + mu_i tmp = 0; + for (i = 0; i < var->cnsts_number; i++) { + tmp += (var->cnsts[i].constraint)->lambda; + } + if (var->bound > 0) + tmp += var->mu; + DEBUG3("\t Working on var (%p). cost = %e; Df = %e", var, tmp, + var->df); + + //uses the partial differential inverse function + tmp = var->func_fpi(var, tmp); - for(i=0; icnsts_number; i++){ - tmp += (var2->cnsts[i].constraint)->lambda; + //computes de overall_error using normalized value + if (overall_error < (fabs(var->value - tmp) / tmp)) { + overall_error = (fabs(var->value - tmp) / tmp); } - if(var2->bound > 0) - tmp += var2->mu; - - if(tmp==0) break; + var->value = tmp; + } + DEBUG3("======> value of var (%p) = %e, overall_error = %e", var, + var->value, overall_error); + } + + if (!__check_kkt(cnst_list, var_list, 0)) + overall_error = 1.0; + DEBUG2("Iteration %d: Overall_error : %f", iteration, overall_error); + } + + + __check_kkt(cnst_list, var_list, 1); + + if (overall_error <= epsilon_min_error) { + DEBUG1("The method converges in %d iterations.", iteration); + } + if (iteration >= max_iterations) { + WARN1 + ("Method reach %d iterations, which is the maximum number of iterations allowed.", + iteration); + } +/* INFO1("Method converged after %d iterations", iteration); */ + + if (XBT_LOG_ISENABLED(surf_lagrange, xbt_log_priority_debug)) { + lmm_print(sys); + } +} + +/* + * Returns a double value corresponding to the result of a dichotomy proccess with + * respect to a given variable/constraint (\mu in the case of a variable or \lambda in + * case of a constraint) and a initial value init. + * + * @param init initial value for \mu or \lambda + * @param diff a function that computes the differential of with respect a \mu or \lambda + * @param var_cnst a pointer to a variable or constraint + * @param min_erro a minimun error tolerated + * + * @return a double correponding to the result of the dichotomyal process + */ +double dichotomy(double init, double diff(double, void *), void *var_cnst, + double min_error) +{ + double min, max; + double overall_error; + double middle; + double min_diff, max_diff, middle_diff; + double diff_0 = 0.0; + min = max = init; + + XBT_IN; + + if (init == 0.0) { + min = max = 0.5; + } - if (tmp==cnst1->lambda) - watch_out=1; - lambda_partial += (-1.0 / tmp); + min_diff = max_diff = middle_diff = 0.0; + overall_error = 1; + + if ((diff_0 = diff(1e-16, var_cnst)) >= 0) { + CDEBUG1(surf_lagrange_dichotomy, "====> returning 0.0 (diff = %e)", + diff_0); + return 0.0; + } + + CDEBUG1(surf_lagrange_dichotomy, + "====> not detected positive diff in 0 (%e)", diff_0); + + while (overall_error > min_error) { + + min_diff = diff(min, var_cnst); + max_diff = diff(max, var_cnst); + + CDEBUG2(surf_lagrange_dichotomy, + "DICHOTOMY ===> min = %1.20f , max = %1.20f", min, max); + CDEBUG2(surf_lagrange_dichotomy, + "DICHOTOMY ===> diffmin = %1.20f , diffmax = %1.20f", min_diff, + max_diff); + + if (min_diff > 0 && max_diff > 0) { + if (min == max) { + CDEBUG0(surf_lagrange_dichotomy, "Decreasing min"); + min = min / 2.0; + } else { + CDEBUG0(surf_lagrange_dichotomy, "Decreasing max"); + max = min; + } + } else if (min_diff < 0 && max_diff < 0) { + if (min == max) { + CDEBUG0(surf_lagrange_dichotomy, "Increasing max"); + max = max * 2.0; + } else { + CDEBUG0(surf_lagrange_dichotomy, "Increasing min"); + min = max; } + } else if (min_diff < 0 && max_diff > 0) { + middle = (max + min) / 2.0; + middle_diff = diff(middle, var_cnst); - if(tmp == 0) - cnst1->new_lambda = LAMBDA_STEP; - else { - lambda_partial += cnst1->bound; - if(watch_out && (lambda_partial>0)) { - /* INFO6("Watch Out (%d) %p! lambda_partial: %e; lambda : %e ; (%e %e) \n",iteration, cnst1, */ - /* lambda_partial, cnst1->lambda, cnst1->lambda / 2, */ - /* cnst1->lambda - sigma_step * lambda_partial); */ - - if(cnst1->lambda < 0) WARN2("Value of cnst1->lambda(%p) = %e < 0", cnst1, cnst1->lambda); - if((cnst1->lambda - sigma_step * lambda_partial) < 0) WARN1("Value of lambda_new = %e < 0", (cnst1->lambda - sigma_step * lambda_partial)); - - if(cnst1->lambda - sigma_step * lambda_partial < cnst1->lambda / 2) - cnst1->new_lambda = cnst1->lambda / 2; - else - cnst1->new_lambda = cnst1->lambda - sigma_step * lambda_partial; - } else - cnst1->new_lambda = cnst1->lambda - sigma_step * lambda_partial; - if(cnst1->new_lambda < 0){ - cnst1->new_lambda = 0; - } + if (max != 0.0 && min != 0.0) { + overall_error = fabs(min - max) / max; } - if(XBT_LOG_ISENABLED(surf_writelambda, xbt_log_priority_debug)) { - fprintf(gnuplot_file, " %e", cnst1->lambda); + if (middle_diff < 0) { + min = middle; + } else if (middle_diff > 0) { + max = middle; + } else { + CWARN0(surf_lagrange_dichotomy, + "Found an optimal solution with 0 error!"); + overall_error = 0; + return middle; } + } else if (min_diff == 0) { + return min; + } else if (max_diff == 0) { + return max; + } else if (min_diff > 0 && max_diff < 0) { + CWARN0(surf_lagrange_dichotomy, + "The impossible happened, partial_diff(min) > 0 && partial_diff(max) < 0"); + } else { + CWARN2(surf_lagrange_dichotomy, + "diffmin (%1.20f) or diffmax (%1.20f) are something I don't know, taking no action.", + min_diff, max_diff); + abort(); } + } + XBT_OUT; - /* - * Now computes the values of each variable (\rho) based on - * the values of \lambda and \mu. - */ - overall_error=0; - xbt_swag_foreach(var1, var_list) { - if(var1->weight <=0) - var1->value = 0.0; - else { - tmp = 0; - for(i=0; icnsts_number; i++){ - tmp += (var1->cnsts[i].constraint)->lambda; - if(var1->bound > 0) - tmp+=var1->mu; - } - - //computes de overall_error - if(overall_error < fabs(var1->value - 1.0/tmp)){ - overall_error = fabs(var1->value - 1.0/tmp); - } + CDEBUG1(surf_lagrange_dichotomy, "====> returning %e", + (min + max) / 2.0); + return ((min + max) / 2.0); +} - var1->value = 1.0 / tmp; - } - - } +/* + * + */ +double partial_diff_mu(double mu, void *param_var) +{ + double mu_partial = 0.0; + double sigma_mu = 0.0; + lmm_variable_t var = (lmm_variable_t) param_var; + int i; + XBT_IN; + //compute sigma_i + for (i = 0; i < var->cnsts_number; i++) + sigma_mu += (var->cnsts[i].constraint)->lambda; + + //compute sigma_i + mu_i + sigma_mu += mu; + + //use auxiliar function passing (sigma_i + mu_i) + mu_partial = diff_aux(var, sigma_mu); + + //add the RTT limit + mu_partial += var->bound; + + XBT_OUT; + return mu_partial; +} +/* + * + */ +double partial_diff_lambda(double lambda, void *param_cnst) +{ - /* Updating lambda's and mu's */ - xbt_swag_foreach(var1, var_list) - if(!((var1->bound > 0.0) || (var1->weight <= 0.0))) - var1->mu = var1->new_mu; - - - xbt_swag_foreach(cnst1, cnst_list) - cnst1->lambda = cnst1->new_lambda; - } + int i; + xbt_swag_t elem_list = NULL; + lmm_element_t elem = NULL; + lmm_variable_t var = NULL; + lmm_constraint_t cnst = (lmm_constraint_t) param_cnst; + double lambda_partial = 0.0; + double sigma_i = 0.0; + XBT_IN; + elem_list = &(cnst->element_set); + CDEBUG1(surf_lagrange_dichotomy,"Computting diff of cnst (%p)", cnst); + xbt_swag_foreach(elem, elem_list) { + var = elem->variable; + if (var->weight <= 0) + continue; - //verify the KKT property - xbt_swag_foreach(cnst1, cnst_list){ - tmp = 0; - elem_list = &(cnst1->element_set); - xbt_swag_foreach(elem1, elem_list) { - var1 = elem1->variable; - if(var1->weight<=0) continue; - tmp += var1->value; + //initilize de sumation variable + sigma_i = 0.0; + + //compute sigma_i of variable var + for (i = 0; i < var->cnsts_number; i++) { + sigma_i += (var->cnsts[i].constraint)->lambda; } - tmp = tmp - cnst1->bound; - + //add mu_i if this flow has a RTT constraint associated + if (var->bound > 0) + sigma_i += var->mu; - if(tmp != 0 || cnst1->lambda != 0){ - WARN4("The link %s(%p) doesn't match the KKT property, value expected (=0) got (lambda=%e) (sum_rho=%e)", (char *)cnst1->id, cnst1, cnst1->lambda, tmp); - } - + //replace value of cnst->lambda by the value of parameter lambda + sigma_i = (sigma_i - cnst->lambda) + lambda; + + //use the auxiliar function passing (\sigma_i + \mu_i) + lambda_partial += diff_aux(var, sigma_i); } - - xbt_swag_foreach(var1, var_list){ - if(var1->bound <= 0 || var1->weight <= 0) continue; - tmp = 0; - tmp = (var1->value - var1->bound); - - if(tmp != 0 || var1->mu != 0){ - WARN4("The flow %s(%p) doesn't match the KKT property, value expected (=0) got (lambda=%e) (sum_rho=%e)", (char *)var1->id, var1, var1->mu, tmp); - } + lambda_partial += cnst->bound; - } + CDEBUG1(surf_lagrange_dichotomy, "returning = %1.20f", lambda_partial); + XBT_OUT; + return lambda_partial; +} - if(overall_error <= epsilon_min_error){ - DEBUG1("The method converge in %d iterations.", iteration); - }else{ - WARN1("Method reach %d iterations, which is the maxmimun number of iterations allowed.", iteration); - } +double diff_aux(lmm_variable_t var, double x) +{ + double tmp_fpi, result; + XBT_IN2("(var (%p), x (%1.20f))", var, x); + xbt_assert0(var->func_fp, + "Initialize the protocol functions first create variables before."); - if(XBT_LOG_ISENABLED(surf_writelambda, xbt_log_priority_debug)) { - fclose(gnuplot_file); - } + tmp_fpi = var->func_fpi(var, x); + result = - tmp_fpi; + CDEBUG1(surf_lagrange_dichotomy, "returning %1.20f", result); + XBT_OUT; + return result; +} +/**************** Vegas and Reno functions *************************/ +/* + * NOTE for Reno: all functions consider the network + * coeficient (alpha) equal to 1. + */ +/* + * For Vegas f: $\alpha_f d_f \log\left(x_f\right)$ + */ +double func_vegas_f(lmm_variable_t var, double x){ + return var->df * log(x); +} + +/* + * For Vegas fp: $\frac{\alpha D_f}{x}$ + */ +double func_vegas_fp(lmm_variable_t var, double x){ + //avoid a disaster value - c'est du bricolage mais ca marche +/* if(x == 0) x = 10e-8; */ + return var->df/x; +} + +/* + * For Vegas fpi: $\frac{\alpha D_f}{x}$ + */ +double func_vegas_fpi(lmm_variable_t var, double x){ + //avoid a disaster value - c'est du bricolage mais ca marche +/* if(x == 0) x = 10e-8; */ + return var->df/x; +} + +/* + * For Vegas fpip: $-\frac{\alpha D_f}{x^2}$ + */ +double func_vegas_fpip(lmm_variable_t var, double x){ + //avoid a disaster value - c'est du bricolage mais ca marche +/* if(x == 0) x = 10e-8; */ + return -( var->df/(x*x) ) ; +} -/* /\* */ -/* * Now computes the values of each variable (\rho) based on */ -/* * the values of \lambda and \mu. */ -/* *\/ */ -/* var_list = &(sys->variable_set); */ -/* xbt_swag_foreach(var1, var_list) { */ -/* tmp = 0; */ -/* for(i=0; icnsts_number; i++){ */ -/* elem1 = &(var1->cnsts[i]); */ -/* tmp += (elem1->constraint)->lambda + var1->mu; */ -/* } */ -/* var1->weight = 1 / tmp; */ -/* DEBUG2("var1->weight (id=%s) : %e", (char *)var1->id, var1->weight); */ -/* } */ +/* + * For Reno f: $\frac{\sqrt{\frac{3}{2}}}{D_f} \arctan\left(\sqrt{\frac{3}{2}}x_f D_f\right)$ + */ +double func_reno_f(lmm_variable_t var, double x){ + xbt_assert0(var->df>0.0,"Don't call me with stupid values!"); + // \sqrt{3/2} = 0.8164965808 + return (0.8164965808 / var->df) * atan( (0.8164965808 / var->df)*x ); +} + +/* + * For Reno fp: $\frac{3}{3 {D_f}^2 x^2 + 2}$ + */ +double func_reno_fp(lmm_variable_t var, double x){ + return 3 / (3*var->df*var->df*x*x + 2); +} +/* + * For Reno fpi: $\sqrt{\frac{1}{{D_f}^2 x} - \frac{2}{3{D_f}^2}}$ + */ +double func_reno_fpi(lmm_variable_t var, double x){ + double res_fpi; + + xbt_assert0(var->df>0.0,"Don't call me with stupid values!"); + xbt_assert0(x>0.0,"Don't call me with stupid values!"); + + res_fpi = 1/(var->df*var->df*x) - 2/(3*var->df*var->df); + if(res_fpi<=0.0) return 0.0; + xbt_assert0(res_fpi>0.0,"Don't call me with stupid values!"); + return sqrt(res_fpi); +} + +/* + * For Reno fpip: $-\frac{1}{2 {D_f}^2 x^2\sqrt{\frac{1}{{D_f}^2 x} - \frac{2}{3{D_f}^2}}}$ + */ +double func_reno_fpip(lmm_variable_t var, double x){ + double res_fpip; + double critical_test; + xbt_assert0(var->df>0.0,"Don't call me with stupid values!"); + xbt_assert0(x>0.0,"Don't call me with stupid values!"); + res_fpip = 1/(var->df*var->df*x) - 2/(3*var->df*var->df); + xbt_assert0(res_fpip>0.0,"Don't call me with stupid values!"); + critical_test = (2*var->df*var->df*x*x*sqrt(res_fpip)); + return -(1.0/critical_test); }