example.c 17.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
/*
 * Copyright (C) 2013 University of Chicago.
 * See COPYRIGHT notice in top-level directory.
 *
 */

/* SUMMARY:
 *
 * This is a sample code to demonstrate CODES usage and best practices.  It
 * sets up a number of servers, each of which is paired up with a simplenet LP
 * to serve as the NIC.  Each server exchanges a sequence of requests and acks
 * with one peer and measures the throughput in terms of payload bytes (ack
 * size) moved per second.
 */

#include <string.h>
#include <assert.h>
#include <ross.h>

#include "codes/lp-io.h"
#include "codes/codes.h"
#include "codes/codes_mapping.h"
#include "codes/configuration.h"
#include "codes/model-net.h"
#include "codes/lp-type-lookup.h"
26
#include "codes/local-storage-model.h"
27

28 29
static int num_reqs = 0;/* number of requests sent by each server (read from config) */
static int payload_sz = 0; /* size of simulated data payload, bytes (read from config) */
30

31 32
/* model-net ID, can be either simple-net, dragonfly or torus (more may be
 * added) */
33 34 35 36
static int net_id = 0;
static int num_servers = 0;
static int offset = 2;

37
/* expected LP group name in configure files for this program */
38
static char *group_name = "SERVERS";
39 40 41 42
/* expected parameter group name for rounds of communication */
static char *param_group_nm = "server_pings";
static char *num_reqs_key = "num_reqs";
static char *payload_sz_key = "payload_sz";
43

44 45 46
typedef struct svr_msg svr_msg;
typedef struct svr_state svr_state;

47
/* types of events that will constitute server activities */
48 49 50 51 52 53 54 55
enum svr_event
{
    KICKOFF,    /* initial event */
    REQ,        /* request event */
    ACK,        /* ack event */
    LOCAL      /* local event */
};

56 57 58
/* this struct serves as the ***persistent*** state of the LP representing the 
 * server in question. This struct is setup when the LP initialization function
 * ptr is called */
59 60 61 62 63 64
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 */
65
    tw_stime end_ts;      /* time that last request finished */
66 67
};

68 69
/* this struct serves as the ***temporary*** event data, which can be thought
 * of as a message between two LPs. */
70 71 72 73 74 75 76 77
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 */
};

78 79 80 81 82 83
/* ROSS expects four functions per LP:
 * - an LP initialization function, called for each LP
 * - an event processing function
 * - a *reverse* event processing function (rollback), and
 * - a finalization/cleanup function when the simulation ends
 */
84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
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);

101 102
/* set up the function pointers for ROSS, as well as the size of the LP state
 * structure (NOTE: ROSS is in charge of event and state (de-)allocation) */
103 104 105 106 107 108 109 110 111 112 113 114 115
tw_lptype svr_lp = {
     (init_f) svr_init,
     (event_f) svr_event,
     (revent_f) svr_rev_event,
     (final_f)  svr_finalize, 
     (map_f) codes_mapping,
     sizeof(svr_state),
};

extern const tw_lptype* svr_get_lp_type();
static void svr_add_lp_type();
static tw_stime ns_to_s(tw_stime ns);
static tw_stime s_to_ns(tw_stime ns);
116 117 118

/* as we only have a single event processing entry point and multiple event
 * types, for clarity we define "handlers" for each (reverse) event type */
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159
static void handle_kickoff_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);
static void handle_ack_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);
static void handle_req_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);
static void handle_local_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
   tw_lp * lp);
static void handle_local_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
   tw_lp * lp);
static void handle_kickoff_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);
static void handle_ack_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);
static void handle_req_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp);

160 161 162 163 164 165 166 167 168 169 170 171
/* for this simulation, each server contacts its neighboring server in an id.
 * this function shows how to use the codes_mapping API to calculate IDs when
 * having to contend with multiple LP types and counts. Note that in this simple
 * example codes_mapping is overkill. */
static tw_lpid get_next_server(tw_lpid sender_id);

/* arguments to be handled by ROSS - strings passed in are expected to be
 * pre-allocated */
static char conf_file_name[256] = {0};
/* this struct contains default parameters used by ROSS, as well as
 * user-specific arguments to be handled by the ROSS config sys. Pass it in
 * prior to calling tw_init */
172 173 174
const tw_optdef app_opt [] =
{
	TWOPT_GROUP("Model net test case" ),
175
        TWOPT_CHAR("codes-config", conf_file_name, "name of codes configuration file"),
176 177 178 179 180 181 182 183 184
	TWOPT_END()
};

int main(
    int argc,
    char **argv)
{
    int nprocs;
    int rank;
185
    int num_nets, *net_ids;
186 187

    /* TODO: explain why we need this (ROSS has cutoff??) */
188 189 190
    g_tw_ts_end = s_to_ns(60*60*24*365); /* one year, in nsecs */

    /* ROSS initialization function calls */
191 192 193
    tw_opt_add(app_opt); /* add user-defined args */
    /* initialize ROSS and parse args. NOTE: tw_init calls MPI_Init */
    tw_init(&argc, &argv); 
194

195
    if (!conf_file_name[0]) 
196
    {
197 198 199
        fprintf(stderr, "Expected \"codes-config\" option, please see --help.\n");
        MPI_Finalize();
        return 1;
200
    }
201

202 203 204
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
  
205 206 207 208 209 210 211 212
    /* loading the config file into the codes-mapping utility, giving us the
     * parsed config object in return. 
     * "config" is a global var defined by codes-mapping */
    if (configuration_load(conf_file_name, MPI_COMM_WORLD, &config)){
        fprintf(stderr, "Error loading config file %s.\n", conf_file_name);
        MPI_Finalize();
        return 1;
    }
213

214 215 216 217 218 219 220 221 222 223 224 225
    /* register model-net LPs with ROSS */
    model_net_register();

    /* register the server LP type with ROSS */
    svr_add_lp_type();

    /* Setup takes the global config object, the registered LPs, and
     * generates/places the LPs as specified in the configuration file.
     * This should only be called after ALL LP types have been registered in 
     * codes */
    codes_mapping_setup();

226
    /* Setup the model-net parameters specified in the global config object,
227 228
     * returned are the identifier(s) for the network type. In this example, we
     * only expect one*/
229
    net_ids = model_net_configure(&num_nets);
230 231 232
    assert(num_nets==1);
    net_id = *net_ids;
    free(net_ids);
233 234 235
    /* in this example, we are using simplenet, which simulates point to point 
     * communication between any two entities (other networks are trickier to
     * setup). Hence: */
236 237 238 239 240 241
    if(net_id != SIMPLENET)
    {
	    printf("\n The test works with simple-net configuration only! ");
	    MPI_Finalize();
	    return 0;
    }
242
    
243 244 245
    /* calculate the number of servers in this simulation,
     * ignoring annotations */
    num_servers = codes_mapping_get_lp_count(group_name, 0, "server", NULL, 1);
246

247 248 249
    /* for this example, we read from a separate configuration group for
     * server message parameters. Since they are constant for all LPs,
     * go ahead and read them prior to running */
250 251
    configuration_get_value_int(&config, param_group_nm, num_reqs_key, NULL, &num_reqs);
    configuration_get_value_int(&config, param_group_nm, payload_sz_key, NULL, &payload_sz);
252

253
    /* begin simulation */ 
254
    tw_run();
255 256

    /* model-net has the capability of outputting network transmission stats */
257 258 259 260 261 262 263 264 265 266 267 268 269
    model_net_report_stats(net_id);

    tw_end();
    return 0;
}

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

static void svr_add_lp_type()
{
270 271 272
    /* lp_type_register should be called exactly once per process per 
     * LP type */
    lp_type_register("server", svr_get_lp_type());
273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
}

static void svr_init(
    svr_state * ns,
    tw_lp * lp)
{
    tw_event *e;
    svr_msg *m;
    tw_stime kickoff_time;
    
    memset(ns, 0, sizeof(*ns));

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

    /* skew each kickoff event slightly to help avoid event ties later on */
    kickoff_time = g_tw_lookahead + tw_rand_unif(lp->rng); 

292
    /* first create the event (time arg is an offset, not absolute time) */
293
    e = codes_event_new(lp->gid, kickoff_time, lp);
294 295
    /* after event is created, grab the allocated message and set msg-specific
     * data */ 
296 297
    m = tw_event_data(e);
    m->svr_event_type = KICKOFF;
298
    /* event is ready to be processed, send it off */
299 300 301 302 303
    tw_event_send(e);

    return;
}

304 305
/* event processing entry point
 * - simply forward the message to the appropriate handler */
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
static void svr_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
   switch (m->svr_event_type)
    {
        case REQ:
            handle_req_event(ns, b, m, lp);
            break;
        case ACK:
            handle_ack_event(ns, b, m, lp);
            break;
        case KICKOFF:
            handle_kickoff_event(ns, b, m, lp);
            break;
	case LOCAL:
	   handle_local_event(ns, b, m, lp); 
	 break;
        default:
	    printf("\n Invalid message type %d ", m->svr_event_type);
            assert(0);
        break;
    }
}

333 334
/* reverse event processing entry point
 * - simply forward the message to the appropriate handler */
335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
static void svr_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
    switch (m->svr_event_type)
    {
        case REQ:
            handle_req_rev_event(ns, b, m, lp);
            break;
        case ACK:
            handle_ack_rev_event(ns, b, m, lp);
            break;
        case KICKOFF:
            handle_kickoff_rev_event(ns, b, m, lp);
            break;
	case LOCAL:
	    handle_local_rev_event(ns, b, m, lp);    
	    break;
        default:
            assert(0);
            break;
    }

    return;
}

363
/* once the simulation is over, do some output */
364 365 366 367
static void svr_finalize(
    svr_state * ns,
    tw_lp * lp)
{
368 369 370 371 372 373 374 375
    printf("server %llu recvd %d bytes in %lf seconds, %lf MiB/s sent_count %d recvd_count %d local_count %d \n", 
            (unsigned long long)(lp->gid/2),
            payload_sz*ns->msg_recvd_count,
            ns_to_s(ns->end_ts-ns->start_ts),
            ((double)(payload_sz*num_reqs)/(double)(1024*1024)/ns_to_s(ns->end_ts-ns->start_ts)),
            ns->msg_sent_count,
            ns->msg_recvd_count,
            ns->local_recvd_count);
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390
    return;
}

/* convert ns to seconds */
static tw_stime ns_to_s(tw_stime ns)
{
    return(ns / (1000.0 * 1000.0 * 1000.0));
}

/* convert seconds to ns */
static tw_stime s_to_ns(tw_stime ns)
{
    return(ns * (1000.0 * 1000.0 * 1000.0));
}

391 392 393 394 395 396 397
/* see declaration for more general info */
tw_lpid get_next_server(tw_lpid sender_id)
{
    tw_lpid rtn_id;
    /* first, get callers LP and group info from codes-mapping. Caching this 
     * info in the LP struct isn't a bad idea for preventing a huge number of
     * lookups */
398 399
    char grp_name[MAX_NAME_LENGTH], lp_type_name[MAX_NAME_LENGTH],
         annotation[MAX_NAME_LENGTH];
400 401
    int  lp_type_id, grp_id, grp_rep_id, offset, num_reps;
    int dest_rep_id;
402 403
    codes_mapping_get_lp_info(sender_id, grp_name, &grp_id, lp_type_name,
            &lp_type_id, annotation, &grp_rep_id, &offset);
404 405 406 407 408 409
    /* in this example, we assume that, for our group of servers, each 
     * "repetition" consists of a single server/NIC pair. Hence, we grab the 
     * server ID for the next repetition, looping around if necessary */
    num_reps = codes_mapping_get_group_reps(grp_name);
    dest_rep_id = (grp_rep_id+1) % num_reps;
    /* finally, get the server (exactly 1 server per rep -> offset w/in rep = 0 */
410 411
    codes_mapping_get_lp_id(grp_name, lp_type_name, NULL, 1, dest_rep_id,
            0, &rtn_id);
412 413 414
    return rtn_id;
}

415 416 417 418 419 420 421
/* handle initial event */
static void handle_kickoff_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
    int dest_id;
    int use_brute_force_map = 0;
    /* normally, when using ROSS, events are allocated as a result of the event
     * creation process. However, since we are now asking model-net to
     * communicate with an entity on our behalf, we need to generate both the
     * message to the recipient and an optional callback message 
     * - thankfully, memory need not persist past the model_net_event call - it
     *   copies the messages */
    svr_msg m_local;
    svr_msg m_remote;

    m_local.svr_event_type = LOCAL;
    m_local.src = lp->gid;
    m_remote.svr_event_type = REQ;
    m_remote.src = lp->gid;
437 438 439 440

    /* record when transfers started on this server */
    ns->start_ts = tw_now(lp);

441 442 443 444 445 446 447 448 449 450 451
    /* each server sends a request to the next highest server 
     * In this simulation, LP determination is simple: LPs are assigned
     * round robin as in serv_1, net_1, serv_2, net_2, etc. 
     * However, that may not always be the case, so we also show a more
     * complicated way to map through codes_mapping */
    if (use_brute_force_map)
        dest_id = (lp->gid + offset)%(num_servers*2);
    else
    {
        dest_id = get_next_server(lp->gid);
    }
452

453
    /* model-net needs to know about (1) higher-level destination LP which is a neighboring server in this case
454
     * (2) struct and size of remote message and (3) struct and size of local message (a local message can be null) */
455
    model_net_event(net_id, "test", dest_id, payload_sz, 0.0, sizeof(svr_msg), 
456
            (const void*)&m_remote, sizeof(svr_msg), (const void*)&m_local, lp);
457 458 459
    ns->msg_sent_count++;
}

460 461
/* at the moment, no need for local callbacks from model-net, so we maintain a
 * count for debugging purposes */ 
462 463 464 465 466 467 468 469 470
static void handle_local_event(
		svr_state * ns,
		tw_bf * b,
		svr_msg * m,
		tw_lp * lp)
{
    ns->local_recvd_count++;
}

471
/* handle recving ack
472 473
 * for this simulation, we repeatedly ping the destination server until num_reqs
 * of size payload_sz have been satisfied - we begin the next req when we
474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490
 * receive an ACK from the destination server */
static void handle_ack_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
    /* the ACK actually doesn't come from the NIC on the other server -
     * model-net "hides" the NIC LP from us so we only see the original
     * destination server */

    /* safety check that this request got to the right server, both with our
     * brute-force lp calculation and our more generic codes-mapping 
     * calculation */
    assert(m->src == (lp->gid + offset)%(num_servers*2) &&
           m->src == get_next_server(lp->gid));

491
    if(ns->msg_sent_count < num_reqs)
492 493 494 495 496 497 498 499 500 501 502
    {
        /* again, allocate our own msgs so model-net can transmit on our behalf */
        svr_msg m_local;
        svr_msg m_remote;

        m_local.svr_event_type = LOCAL;
        m_local.src = lp->gid;
        m_remote.svr_event_type = REQ;
        m_remote.src = lp->gid;

        /* send another request */
503
	model_net_event(net_id, "test", m->src, payload_sz, 0.0, sizeof(svr_msg), 
504 505 506 507 508 509 510 511 512
                (const void*)&m_remote, sizeof(svr_msg), (const void*)&m_local, lp);
        ns->msg_sent_count++;
        m->incremented_flag = 1;
        
    }
    else
    {
	/* threshold count reached, stop sending messages */
        m->incremented_flag = 0;
513
        ns->end_ts = tw_now(lp);
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543
    }
    return;
}

/* handle receiving request */
static void handle_req_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
    svr_msg m_local;
    svr_msg m_remote;

    m_local.svr_event_type = LOCAL;
    m_local.src = lp->gid;
    m_remote.svr_event_type = ACK;
    m_remote.src = lp->gid;

    /* safety check that this request got to the right server */
    
    assert(lp->gid == (m->src + offset)%(num_servers*2) &&
           lp->gid == get_next_server(m->src));
    ns->msg_recvd_count++;

    /* send ack back */
    /* simulated payload of 1 MiB */
    /* also trigger a local event for completion of payload msg */
    /* remote host will get an ack event */
   
544
    model_net_event(net_id, "test", m->src, payload_sz, 0.0, sizeof(svr_msg), 
545 546 547 548 549 550 551 552 553
            (const void*)&m_remote, sizeof(svr_msg), (const void*)&m_local, lp);
    return;
}

/* for us, reverse events are very easy, the only LP state that needs to be
 * rolled back are the counts.
 * for more complex simulations, this will not be the case (e.g., state
 * containing queues) */

554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569
static void handle_local_rev_event(
	       svr_state * ns,
	       tw_bf * b,
	       svr_msg * m,
	       tw_lp * lp)
{
   ns->local_recvd_count--;
}
/* reverse handler for req event */
static void handle_req_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
    ns->msg_recvd_count--;
570
    /* model-net has its own reverse computation support */ 
571
    model_net_event_rc(net_id, lp, payload_sz);
572 573 574 575 576 577 578 579 580 581 582 583 584

    return;
}


/* reverse handler for kickoff */
static void handle_kickoff_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
    ns->msg_sent_count--;
585
    model_net_event_rc(net_id, lp, payload_sz);
586 587 588 589 590 591 592 593 594 595 596 597 598

    return;
}

/* reverse handler for ack*/
static void handle_ack_rev_event(
    svr_state * ns,
    tw_bf * b,
    svr_msg * m,
    tw_lp * lp)
{
    if(m->incremented_flag)
    {
599
        model_net_event_rc(net_id, lp, payload_sz);
600 601 602 603 604 605 606 607 608 609 610 611 612
        ns->msg_sent_count--;
    }
    return;
}

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