model-net-synthetic-fattree.c 15.4 KB
Newer Older
1 2 3 4 5 6 7 8
/*
 * Copyright (C) 2015 University of Chicago.
 * See COPYRIGHT notice in top-level directory.
 *
 */

/*
* The test program generates some synthetic traffic patterns for the model-net network models.
9
* currently it only support the fat tree network model uniform random and nearest neighbor traffic patterns.
10 11 12 13
*/

#include "codes/model-net.h"
#include "codes/lp-io.h"
14
#include "codes/net/fattree.h"
15 16 17 18 19 20 21
#include "codes/codes.h"
#include "codes/codes_mapping.h"
#include "codes/configuration.h"
#include "codes/lp-type-lookup.h"

#define PAYLOAD_SZ 512

22 23
#define PARAMS_LOG 1

24 25 26 27
static int net_id = 0;
static int offset = 2;
static int traffic = 1;
static double arrival_time = 1000.0;
28 29
static double load = 0.0;	//Percent utilization of terminal uplink
static double MEAN_INTERVAL = 0.0;
30
char * modelnet_stats_dir;
31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
/* whether to pull instead of push */

static int num_servers_per_rep = 0;
static int num_routers_per_grp = 0;
static int num_nodes_per_grp = 0;

static int num_groups = 0;
static int num_nodes = 0;

typedef struct svr_msg svr_msg;
typedef struct svr_state svr_state;

/* global variables for codes mapping */
static char group_name[MAX_NAME_LENGTH];
static char lp_type_name[MAX_NAME_LENGTH];
static int group_index, lp_type_index, rep_id, offset;

48 49 50 51 52 53 54 55 56 57 58 59 60 61 62
/* convert GiB/s and bytes to ns */
 static tw_stime bytes_to_ns(uint64_t bytes, double GB_p_s)
 {
     tw_stime time;
 
     /* bytes to GB */
     time = ((double)bytes)/(1024.0*1024.0*1024.0);
     /* GiB to s */
     time = time / GB_p_s;
     /* s to ns */
     time = time * 1000.0 * 1000.0 * 1000.0;
 
     return(time);
 }

63 64 65 66 67 68 69 70 71 72 73 74
/* type of events */
enum svr_event
{
    KICKOFF,	   /* kickoff event */
    REMOTE,        /* remote event */
    LOCAL      /* local event */
};

/* type of synthetic traffic */
enum TRAFFIC
{
	UNIFORM = 1, /* sends message to a randomly selected node */
75 76 77
    BISECTION = 2, /* sends messages to node established in bisection pairing*/
	NEAREST_GROUP = 3, /* sends message to the node connected to the neighboring router */
	NEAREST_NEIGHBOR = 4 /* sends message to the next node (potentially connected to the same router) */
78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93
};

struct svr_state
{
    int msg_sent_count;   /* requests sent */
    int msg_recvd_count;  /* requests recvd */
    int local_recvd_count; /* number of local messages received */
    tw_stime start_ts;    /* time that we started sending requests */
    tw_stime end_ts;      /* time that we ended sending requests */
};

struct svr_msg
{
    enum svr_event svr_event_type;
    tw_lpid src;          /* source of this request or ack */
    int incremented_flag; /* helper for reverse computation */
94
    model_net_event_return event_rc;
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
};

static void svr_init(
    svr_state * ns,
    tw_lp * lp);
static void svr_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);
static void svr_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);
static void svr_finalize(
    svr_state * ns,
    tw_lp * lp);

tw_lptype svr_lp = {
    (init_f) svr_init,
    (pre_run_f) NULL,
    (event_f) svr_event,
    (revent_f) svr_rev_event,
119
    (commit_f) NULL,
120 121 122 123 124
    (final_f)  svr_finalize,
    (map_f) codes_mapping,
    sizeof(svr_state),
};

125 126
/* setup for the ROSS event tracing
 */
127
void ft_svr_event_collect(svr_msg *m, tw_lp *lp, char *buffer, int *collect_flag)
128
{
Misbah Mubarak's avatar
Misbah Mubarak committed
129 130 131 132
    (void)lp;
    (void)buffer;
    (void)collect_flag;

133 134 135 136
    int type = (int) m->svr_event_type;
    memcpy(buffer, &type, sizeof(type));
}

137 138 139 140
/* can add in any model level data to be collected along with simulation engine data
 * in the ROSS instrumentation.  Will need to update the last field in 
 * ft_svr_model_types[0] for the size of the data to save in each function call
 */
141 142
void ft_svr_model_stat_collect(svr_state *s, tw_lp *lp, char *buffer)
{
Misbah Mubarak's avatar
Misbah Mubarak committed
143 144 145 146
    (void)s;
    (void)lp;
    (void)buffer;

147 148 149 150
    return;
}

st_model_types ft_svr_model_types[] = {
151
    {(ev_trace_f) ft_svr_event_collect,
152 153
     sizeof(int),
     (model_stat_f) ft_svr_model_stat_collect,
154
     0,
155 156 157 158
     NULL,
     NULL,
     0},
    {NULL, 0, NULL, 0, NULL, NULL, 0}
159 160
};

161
static const st_model_types  *ft_svr_get_model_stat_types(void)
162
{
163
    return(&ft_svr_model_types[0]);
164 165
}

166
void ft_svr_register_model_stats()
167
{
168
    st_model_type_register("server", ft_svr_get_model_stat_types());
169 170
}

171 172 173
const tw_optdef app_opt [] =
{
        TWOPT_GROUP("Model net synthetic traffic " ),
174
	TWOPT_UINT("traffic", traffic, "UNIFORM RANDOM=1, BISECTION=2 "),
175
	TWOPT_STIME("arrival_time", arrival_time, "INTER-ARRIVAL TIME"),
176
        TWOPT_STIME("load", load, "percentage of terminal link bandiwdth to inject packets"),
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193
        TWOPT_END()
};

const tw_lptype* svr_get_lp_type()
{
            return(&svr_lp);
}

static void svr_add_lp_type()
{
  lp_type_register("server", svr_get_lp_type());
}

static void issue_event(
    svr_state * ns,
    tw_lp * lp)
{
194
    (void)ns;
195 196 197 198 199 200 201 202
    tw_event *e;
    svr_msg *m;
    tw_stime kickoff_time;

    /* each server sends a dummy event to itself that will kick off the real
     * simulation
     */

203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
    int this_packet_size = 0;
    double this_link_bandwidth = 0.0;

    configuration_get_value_int(&config, "PARAMS", "packet_size", NULL, &this_packet_size);
    if(!this_packet_size) {
        this_packet_size = 0;
        fprintf(stderr, "Packet size not specified, setting to %d\n", this_packet_size);
        exit(0);
    }

    configuration_get_value_double(&config, "PARAMS", "link_bandwidth", NULL, &this_link_bandwidth);
    if(!this_link_bandwidth) {
        this_link_bandwidth = 4.7;
        fprintf(stderr, "Bandwidth of channels not specified, setting to %lf\n", this_link_bandwidth);
    }

    if(arrival_time!=0)
    {
        MEAN_INTERVAL = arrival_time;
    }
    if(load != 0)
    {
        MEAN_INTERVAL = bytes_to_ns(this_packet_size, load*this_link_bandwidth);
    }

228
    /* skew each kickoff event slightly to help avoid event ties later on */
229 230
//    kickoff_time = 1.1 * g_tw_lookahead + tw_rand_exponential(lp->rng, arrival_time);
    kickoff_time = g_tw_lookahead + tw_rand_exponential(lp->rng, MEAN_INTERVAL);
231 232 233 234 235 236 237 238 239 240 241

    e = tw_event_new(lp->gid, kickoff_time, lp);
    m = tw_event_data(e);
    m->svr_event_type = KICKOFF;
    tw_event_send(e);
}

static void svr_init(
    svr_state * ns,
    tw_lp * lp)
{
242 243
    ns->start_ts = 0.0;

244 245 246 247 248 249 250 251 252 253
    issue_event(ns, lp);
    return;
}

static void handle_kickoff_rev_event(
            svr_state * ns,
            tw_bf * b,
            svr_msg * m,
            tw_lp * lp)
{
254 255
    (void)b;
    (void)m;
256
	ns->msg_sent_count--;
257
	model_net_event_rc2(lp, &m->event_rc);
258
    tw_rand_reverse_unif(lp->rng);
259 260 261 262 263 264 265
}	
static void handle_kickoff_event(
	    svr_state * ns,
	    tw_bf * b,
	    svr_msg * m,
	    tw_lp * lp)
{
266 267
    (void)b;
    (void)m;
268 269
//    char* anno;
    char anno[MAX_NAME_LENGTH];
270 271 272 273 274 275 276 277 278 279 280 281
    tw_lpid local_dest = -1, global_dest = -1;
   
    svr_msg * m_local = malloc(sizeof(svr_msg));
    svr_msg * m_remote = malloc(sizeof(svr_msg));

    m_local->svr_event_type = LOCAL;
    m_local->src = lp->gid;

    memcpy(m_remote, m_local, sizeof(svr_msg));
    m_remote->svr_event_type = REMOTE;

    ns->start_ts = tw_now(lp);
282

283
   codes_mapping_get_lp_info(lp->gid, group_name, &group_index, lp_type_name, &lp_type_index, anno, &rep_id, &offset);
284
   int local_id = codes_mapping_get_lp_relative_id(lp->gid, 0, 0);
285 286 287 288 289
   /* in case of uniform random traffic, send to a random destination. */
   if(traffic == UNIFORM)
   {
   	local_dest = tw_rand_integer(lp->rng, 0, num_nodes - 1);
   }
290 291 292 293
   if(traffic == BISECTION)
   {
        local_dest = (local_id + num_nodes/2) % num_nodes;
   }
294

295
   assert(local_dest < LLU(num_nodes));
296 297 298

   global_dest = codes_mapping_get_lpid_from_relative(local_dest, group_name, lp_type_name, NULL, 0);

299 300
   //printf("global_src:%d, local_src:%d, global_dest:%d, local_dest:%d num_nodes:%d \n",(int)lp->gid, local_id, (int)global_dest,(int)local_dest, num_nodes);

301 302
   // If Destination is self, then generate new destination
   if((int)global_dest == (int)lp->gid)
303
   {
304 305
       local_dest = (local_dest+1) % (num_nodes-1);
       global_dest = codes_mapping_get_lpid_from_relative(local_dest, group_name, lp_type_name, NULL, 0);
306
   }
307

308
   ns->msg_sent_count++;
309

310
   m->event_rc = model_net_event(net_id, "test", global_dest, PAYLOAD_SZ, 0.0, sizeof(svr_msg), (const void*)m_remote, sizeof(svr_msg), (const void*)m_local, lp);
311 312

   //printf("LP:%d localID:%d Here\n",(int)lp->gid, (int)local_dest);
313
   issue_event(ns, lp);
314
   //printf("Just Checking net_id:%d\n",net_id);
315 316 317 318 319 320 321 322 323
   return;
}

static void handle_remote_rev_event(
            svr_state * ns,     
            tw_bf * b,
            svr_msg * m,
            tw_lp * lp)
{
324 325 326
        (void)b;
        (void)m;
        (void)lp;
327 328 329 330 331 332 333 334 335
        ns->msg_recvd_count--;
}

static void handle_remote_event(
	    svr_state * ns,
	    tw_bf * b,
	    svr_msg * m,
	    tw_lp * lp)
{
336 337 338
    (void)b;
    (void)m;
    (void)lp;
339 340 341 342 343 344 345 346 347
	ns->msg_recvd_count++;
}

static void handle_local_rev_event(
                svr_state * ns,
                tw_bf * b,
                svr_msg * m,
                tw_lp * lp)
{
348 349 350
    (void)b;
    (void)m;
    (void)lp;
351 352 353 354 355 356 357 358 359
	ns->local_recvd_count--;
}

static void handle_local_event(
                svr_state * ns,
                tw_bf * b,
                svr_msg * m,
                tw_lp * lp)
{
360 361 362
    (void)b;
    (void)m;
    (void)lp;
363 364 365 366 367 368 369 370 371
    ns->local_recvd_count++;
}

static void svr_finalize(
    svr_state * ns,
    tw_lp * lp)
{
    ns->end_ts = tw_now(lp);

372 373
//    printf("server %llu recvd %d bytes in %f seconds, %f MiB/s sent_count %d recvd_count %d local_count %d \n", (unsigned long long)lp->gid, PAYLOAD_SZ*ns->msg_recvd_count, ns_to_s(ns->end_ts-ns->start_ts),
//        ((double)(PAYLOAD_SZ*ns->msg_sent_count)/(double)(1024*1024)/ns_to_s(ns->end_ts-ns->start_ts)), ns->msg_sent_count, ns->msg_recvd_count, ns->local_recvd_count);
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417
    return;
}

static void svr_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
    switch (m->svr_event_type)
    {
	case REMOTE:
		handle_remote_rev_event(ns, b, m, lp);
		break;
	case LOCAL:
		handle_local_rev_event(ns, b, m, lp);
		break;
	case KICKOFF:
		handle_kickoff_rev_event(ns, b, m, lp);
		break;
	default:
		assert(0);
		break;
    }
}

static void svr_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
   switch (m->svr_event_type)
    {
        case REMOTE:
            handle_remote_event(ns, b, m, lp);
            break;
        case LOCAL:
            handle_local_event(ns, b, m, lp);
            break;
	case KICKOFF:
	    handle_kickoff_event(ns, b, m, lp);
	    break;
        default:
418
            printf("\n LP: %d has received invalid message from src lpID: %d of message type:%d", (int)lp->gid, (int)m->src, m->svr_event_type);
419 420 421 422 423 424 425 426 427
            assert(0);
        break;
    }
}

int main(
    int argc,
    char **argv)
{
428

429 430 431 432 433 434 435 436
    int nprocs;
    int rank;
    int num_nets;
    int *net_ids;

    lp_io_handle handle;

    tw_opt_add(app_opt);
437

438
    tw_init(&argc, &argv);
439

440 441 442 443 444 445 446 447 448 449 450 451 452
    offset = 1;

    if(argc < 2)
    {
            printf("\n Usage: mpirun <args> --sync=2/3 mapping_file_name.conf (optional --nkp) ");
            MPI_Finalize();
            return 0;
    }

    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    MPI_Comm_size(MPI_COMM_WORLD, &nprocs);

    configuration_load(argv[2], MPI_COMM_WORLD, &config);
453

454
    model_net_register();
455

456
    svr_add_lp_type();
457

458
    if (g_st_ev_trace || g_st_model_stats || g_st_use_analysis_lps)
459
        ft_svr_register_model_stats();
460

461
    codes_mapping_setup();
462

463 464

    net_ids = model_net_configure(&num_nets);
465
    //assert(num_nets==1);
466 467 468
    net_id = *net_ids;
    free(net_ids);

469
    if(net_id != FATTREE)
470
    {
471
	printf("\n The test works with fat tree model configuration only! ");
472 473 474 475 476
        MPI_Finalize();
        return 0;
    }
    num_servers_per_rep = codes_mapping_get_lp_count("MODELNET_GRP", 1, "server",
            NULL, 1);
477
    configuration_get_value_int(&config, "PARAMS", "num_routers", NULL, &num_routers_per_grp);
478 479 480 481 482
    
    num_groups = (num_routers_per_grp * (num_routers_per_grp/2) + 1);
    num_nodes = num_groups * num_routers_per_grp * (num_routers_per_grp / 2);
    num_nodes_per_grp = num_routers_per_grp * (num_routers_per_grp / 2);

483 484
    num_nodes = codes_mapping_get_lp_count("MODELNET_GRP", 0, "server", NULL, 1);

485
    printf("num_nodes:%d \n",num_nodes);
486

487 488 489 490
    if(lp_io_prepare("modelnet-test", LP_IO_UNIQ_SUFFIX, &handle, MPI_COMM_WORLD) < 0)
    {
        return(-1);
    }
491
    modelnet_stats_dir = lp_io_handle_to_dir(handle);
492 493

    tw_run();
494

495 496
    model_net_report_stats(net_id);

497 498 499 500 501 502 503 504 505 506 507 508 509 510 511

#if PARAMS_LOG
    if(!g_tw_mynode)
    {
	char temp_filename[1024];
	char temp_filename_header[1024];
	sprintf(temp_filename,"%s/sim_log.txt",modelnet_stats_dir);
	sprintf(temp_filename_header,"%s/sim_log_header.txt",modelnet_stats_dir);
	FILE *fattree_results_log=fopen(temp_filename, "a");
	FILE *fattree_results_log_header=fopen(temp_filename_header, "a");
	if(fattree_results_log == NULL)
		printf("\n Failed to open results log file %s in synthetic-fattree\n",temp_filename);
	if(fattree_results_log_header == NULL)
		printf("\n Failed to open results log header file %s in synthetic-fattree\n",temp_filename_header);
	printf("Printing Simulation Parameters/Results Log File\n");
512 513
	fprintf(fattree_results_log_header,", <Workload>, <Load>, <Mean Interval>, ");
	fprintf(fattree_results_log,"%11.3d, %5.2f, %15.2f, ",traffic, load, MEAN_INTERVAL);
514 515 516 517 518
	fclose(fattree_results_log_header);
	fclose(fattree_results_log);
    }
#endif

519 520 521 522 523 524
    if(lp_io_flush(handle, MPI_COMM_WORLD) < 0)
    {
        return(-1);
    }

    tw_end();
525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587

#if PARAMS_LOG
    if(!g_tw_mynode)
    {
	char temp_filename[1024];
	char temp_filename_header[1024];
	sprintf(temp_filename,"%s/sim_log.txt",modelnet_stats_dir);
	sprintf(temp_filename_header,"%s/sim_log_header.txt",modelnet_stats_dir);
	FILE *fattree_results_log=fopen(temp_filename, "a");
	FILE *fattree_results_log_header=fopen(temp_filename_header, "a");
	FILE *fattree_ross_csv_log=fopen("ross.csv", "r");
	if(fattree_results_log == NULL)
		printf("\n Failed to open results log file %s in synthetic-fattree\n",temp_filename);
	if(fattree_results_log_header == NULL)
		printf("\n Failed to open results log header file %s in synthetic-fattree\n",temp_filename_header);
	if(fattree_ross_csv_log == NULL)
		tw_error(TW_LOC, "\n Failed to open ross.csv log file \n");
	printf("Reading ROSS specific data from ross.csv and Printing to Fat Tree Log File\n");
	
	char * line = NULL;
	size_t len = 0;
	ssize_t read = getline(&line, &len, fattree_ross_csv_log);
	while (read != -1) 
	{
		read = getline(&line, &len, fattree_ross_csv_log);
	}

	char * pch;
	pch = strtok (line,",");
	int idx = 0;
        int gvt_computations;
	long long total_events, rollbacks, net_events;
        float running_time, efficiency, event_rate;
	while (pch != NULL)
	{
		pch = strtok (NULL, ",");
		switch(idx)
		{
			case 4:
				total_events = atoll(pch);
				break;
			case 13:
				rollbacks = atoll(pch);
				break;
			case 17:
				gvt_computations = atoi(pch);
				break;
			case 18:
				net_events = atoll(pch);
				break;
			case 3:
				running_time = atof(pch);
				break;
			case 8:
				efficiency = atof(pch);
				break;
			case 19:
				event_rate = atof(pch);
				break;
		}
		idx++;
	}
	fprintf(fattree_results_log_header,"<Total Events>, <Rollbacks>, <GVT Computations>, <Net Events>, <Running Time>, <Efficiency>, <Event Rate>");
588
	fprintf(fattree_results_log,"%14llu, %11llu, %18d, %12llu, %14.4f, %12.2f, %12.2f\n",total_events,rollbacks,gvt_computations,net_events,running_time,efficiency,event_rate);
589 590 591 592 593 594
	fclose(fattree_results_log);
	fclose(fattree_results_log_header);
	fclose(fattree_ross_csv_log);
    }
#endif

595 596 597 598 599 600 601 602 603 604 605
    return 0;
}

/*
 * Local variables:
 *  c-indent-level: 4
 *  c-basic-offset: 4
 * End:
 *
 * vim: ft=c ts=8 sts=4 sw=4 expandtab
 */