vitalif
/
mdtest
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fixed striding/segmentation in N:N and N:1 cases, for multi-part-upload. 

All ranks locally capture and accumulate Etags for the parts they are
writing.  In the N:1 cases, these are ethen collected by rank 0, via
MPI_Gather.  This is effectively an organization matching the "segmented"
layout.  If data was written segmented, then rank0 assigns part-numbers to
with appropriate offsets to correspond to what would've been used by each
rank when writing a given etag.  If data was written strided, then etags
must also be accessed in strided order, to build the XML that will be sent.

TBD: Once the total volume of etag data exceeds the size of memory at rank
0, we'll need to impose a more-sophisticated technique.  One idea is to
thread the MPI comms differently from the libcurl comms, so that multiple
gathers can be staged incrementally, while sending a single stream of XML
data tot he servers.  For example, the libcurl write-function could
interact with the MPI prog to allow the appearance of a single stream of
data.
master
Jeffrey Thornton Inman 2014-12-02 09:10:32 -07:00
parent b26f308191
commit 4368cc2dc4
2 changed files with 370 additions and 215 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

579
src/aiori-S3.c
View File

@ -64,17 +64,7 @@
* --> Select this option with the '-a S3' command-line arg to IOR * --> Select this option with the '-a S3' command-line arg to IOR
* *
* *
* (2) "S3 + EMC append" uses S3 Multi-Part Upload for N:1, like pure S3, * (2) "EMC S3 Extensions" uses the EMC byte-range support for N:1
* but also allows appends in the N:N case, via the EMC byte-range
* write support. This also does away with constraints on the number
* or size of parts to S3 Multi-Part Upload.
*
* --> Select this option with the '-a S3_plus' command-line arg to IOR.
* ior.c will then set the IOR_S3_EMC_EXT flag, which will cause
* s3_connect to initialize with EMC-extensions enabled.
*
*
* (3) "EMC S3 Extensions" uses the EMC byte-range support for N:1
* writes, eliminating Multi-Part Upload. EMC expects this will * writes, eliminating Multi-Part Upload. EMC expects this will
* perform better than MPU, and it avoids some problems that are * perform better than MPU, and it avoids some problems that are
* imposed by the S3 MPU spec. [See comments at EMC_Xfer().] * imposed by the S3 MPU spec. [See comments at EMC_Xfer().]
@ -84,10 +74,10 @@
* *
* NOTE: Putting EMC's S3-extensions in the same file with the S3 API * NOTE: Putting EMC's S3-extensions in the same file with the S3 API
* allows us to share some code that would otherwise be duplicated * allows us to share some code that would otherwise be duplicated
* (e.g. s3_connect(), etc). This should help us avoid losing bug * (e.g. s3_connect(), etc). This should also help us avoid losing
* fixes that are discovered in one interface or the other. In some * bug fixes that are discovered in one interface or the other. In
* cases, S3 is incapable of supporting all the needs of IOR. (For * some cases, S3 is incapable of supporting all the needs of IOR.
* example, see notes about "append", above S3_Xfer(). * (For example, see notes about "append", above S3_Xfer().
* *
******************************************************************************/ ******************************************************************************/
@ -283,6 +273,7 @@ s3_connect( IOR_param_t* param ) {
aws_iobuf_growth_size(param->etags, 1024*1024*8); aws_iobuf_growth_size(param->etags, 1024*1024*8);
// our hosts are currently 10.140.0.15 - 10.140 0.18 // our hosts are currently 10.140.0.15 - 10.140 0.18
// TBD: Try DNS-round-robin server at vi-lb.ccstar.lanl.gov
snprintf(buff, BUFF_SIZE, "10.140.0.%d:9020", 15 + (rank % 4)); snprintf(buff, BUFF_SIZE, "10.140.0.%d:9020", 15 + (rank % 4));
s3_set_host(buff); s3_set_host(buff);
@ -347,16 +338,11 @@ s3_MPU_reset(IOR_param_t* param) {
/* /*
* One doesn't "open" an object, in REST semantics. All we really care * One doesn't "open" an object, in REST semantics. All we really care
* about is whether caller expects the object to have zero-size, when we * about is whether caller expects the object to have zero-size, when we
* return. If so, we have to delete it, then recreate it empty. * return. If so, we conceptually delete it, then recreate it empty.
*
* NOTE: Similarly, there's no file-descriptor to return. On the other
* hand, we keep needing the file *NAME*. Therefore, we will return
* the file-name, and let IOR pass it around to our functions, in
* place of its usual file-descriptor argument.
* *
* ISSUE: If the object is going to receive "appends" (supported in EMC S3 * ISSUE: If the object is going to receive "appends" (supported in EMC S3
* extensions), the object has to exist before the first append * extensions), the object has to exist before the first append
* operation. On the other hand, There appears to be a bug in the * operation. On the other hand, there appears to be a bug in the
* EMC implementation, such that if an object ever receives appends, * EMC implementation, such that if an object ever receives appends,
* and then is deleted, and then recreated, the recreated object will * and then is deleted, and then recreated, the recreated object will
* always return "500 Server Error" on GET (whether it has been * always return "500 Server Error" on GET (whether it has been
@ -364,6 +350,13 @@ s3_MPU_reset(IOR_param_t* param) {
* *
* Therefore, a safer thing to do here is write zero-length contents, * Therefore, a safer thing to do here is write zero-length contents,
* instead of deleting. * instead of deleting.
*
* NOTE: There's also no file-descriptor to return, in REST semantics. On
* the other hand, we keep needing the file *NAME*. Therefore, we
* will return the file-name, and let IOR pass it around to our
* functions, in place of what IOR understands to be a
* file-descriptor.
*
*/ */
static static
@ -374,7 +367,8 @@ S3_Create_Or_Open_internal(char* testFileName,
int multi_part_upload_p ) { int multi_part_upload_p ) {
if (param->verbose >= VERBOSE_2) { if (param->verbose >= VERBOSE_2) {
printf("-> S3_Create_Or_Open\n"); printf("-> S3_Create_Or_Open('%s', ,%d, %d)\n",
testFileName, createFile, multi_part_upload_p);
} }
/* initialize curl, if needed */ /* initialize curl, if needed */
@ -388,46 +382,38 @@ S3_Create_Or_Open_internal(char* testFileName,
fprintf( stdout, "Direct I/O mode is not implemented in S3\n" ); fprintf( stdout, "Direct I/O mode is not implemented in S3\n" );
} }
// easier to think
int n_to_n = param->filePerProc;
int n_to_1 = ! n_to_n;
/* check whether object needs reset to zero-length */ /* check whether object needs reset to zero-length */
int needs_reset = 0; int needs_reset = 0;
if (! multi_part_upload_p) if (! multi_part_upload_p)
needs_reset = 1; /* so "append" can work */ needs_reset = 1; /* so "append" can work */
if ( param->openFlags & IOR_TRUNC ) else if ( param->openFlags & IOR_TRUNC )
needs_reset = 1; /* so "append" can work */ needs_reset = 1; /* so "append" can work */
else if (createFile) { else if (createFile) {
AWS4C_CHECK( s3_head(param->io_buf, testFileName) ); // AWS4C_CHECK( s3_head(param->io_buf, testFileName) );
if ( ! AWS4C_OK(param->io_buf) ) // if ( ! AWS4C_OK(param->io_buf) )
needs_reset = 1; needs_reset = 1;
} }
if ( param->open == WRITE ) { if ( param->open == WRITE ) {
/* initializations for N:N writes */ /* initializations for N:1 or N:N writes using multi-part upload */
if ( param->filePerProc ) { if (multi_part_upload_p) {
/* maybe reset to zero-length, so "append" can work */ // For N:N, all ranks do their own MPU open/close. For N:1, only
if (needs_reset) { // rank0 does that. Either way, the response from the server
aws_iobuf_reset(param->io_buf); // includes an "uploadId", which must be used to upload parts to
AWS4C_CHECK( s3_put(param->io_buf, testFileName) ); // the same object.
AWS4C_CHECK_OK( param->io_buf ); if ( n_to_n || (rank == 0) ) {
}
}
/* initializations for N:1 writes using multi-part upload */
else if (multi_part_upload_p) {
/* rank0 initiates multi-part upload. The response from the server
includes an "uploadId", which must be used by all ranks, when
uploading parts. */
if (rank == 0) {
// rank0 handles truncate // rank0 handles truncate
if ( needs_reset) { if ( needs_reset) {
aws_iobuf_reset(param->io_buf); aws_iobuf_reset(param->io_buf);
AWS4C_CHECK( s3_put(param->io_buf, testFileName) ); AWS4C_CHECK( s3_put(param->io_buf, testFileName) ); /* 0-length write */
AWS4C_CHECK_OK( param->io_buf );
} }
// POST request with URL+"?uploads" initiates multi-part upload // POST request with URL+"?uploads" initiates multi-part upload
@ -450,7 +436,7 @@ S3_Create_Or_Open_internal(char* testFileName,
if (! upload_id) if (! upload_id)
ERR_SIMPLE("couldn't find 'UploadId' in returned XML\n"); ERR_SIMPLE("couldn't find 'UploadId' in returned XML\n");
if (param->verbose >= VERBOSE_4) if (param->verbose >= VERBOSE_3)
printf("got UploadId = '%s'\n", upload_id); printf("got UploadId = '%s'\n", upload_id);
const size_t upload_id_len = strlen(upload_id); const size_t upload_id_len = strlen(upload_id);
@ -469,19 +455,26 @@ S3_Create_Or_Open_internal(char* testFileName,
xmlFreeDoc(doc); xmlFreeDoc(doc);
aws_iobuf_free(response); aws_iobuf_free(response);
// share UploadId across all ranks // For N:1, share UploadId across all ranks
MPI_Bcast(param->UploadId, MAX_UPLOAD_ID_SIZE, MPI_BYTE, 0, param->testComm); if (n_to_1)
MPI_Bcast(param->UploadId, MAX_UPLOAD_ID_SIZE, MPI_BYTE, 0, param->testComm);
} }
else else
// recv UploadID from Rank 0 // N:1, and we're not rank0. recv UploadID from Rank 0
MPI_Bcast(param->UploadId, MAX_UPLOAD_ID_SIZE, MPI_BYTE, 0, param->testComm); MPI_Bcast(param->UploadId, MAX_UPLOAD_ID_SIZE, MPI_BYTE, 0, param->testComm);
} }
/* initializations for N:1 writes using EMC byte-range extensions */ /* initializations for N:N or N:1 writes using EMC byte-range extensions */
else { else {
/* maybe reset to zero-length, so "append" can work */ /* maybe reset to zero-length, so "append" can work */
if (needs_reset) { if (needs_reset) {
if (verbose >= VERBOSE_3) {
fprintf( stdout, "rank %d resetting\n",
rank);
}
aws_iobuf_reset(param->io_buf); aws_iobuf_reset(param->io_buf);
AWS4C_CHECK( s3_put(param->io_buf, testFileName) ); AWS4C_CHECK( s3_put(param->io_buf, testFileName) );
AWS4C_CHECK_OK( param->io_buf ); AWS4C_CHECK_OK( param->io_buf );
@ -662,47 +655,67 @@ S3_Xfer_internal(int access,
char* data_ptr = (char *)buffer; char* data_ptr = (char *)buffer;
off_t offset = param->offset; off_t offset = param->offset;
// easier to think
int n_to_n = param->filePerProc;
int n_to_1 = (! n_to_n);
int segmented = (param->segmentCount == 1);
if (access == WRITE) { /* WRITE */ if (access == WRITE) { /* WRITE */
if (verbose >= VERBOSE_4) { if (verbose >= VERBOSE_3) {
fprintf( stdout, "task %d writing to offset %lld\n", fprintf( stdout, "rank %d writing length=%lld to offset %lld\n",
rank, rank,
remaining,
param->offset + length - remaining); param->offset + length - remaining);
} }
if (param->filePerProc) { // N:N if (multi_part_upload_p) {
// NOTE: You must call 's3_enable_EMC_extensions(1)' to let this work: // For N:1, part-numbers must have a global ordering for the
s3_set_byte_range(-1,-1); // produces header "Range: bytes=-1-" // components of the final object. param->part_number is
// incremented by 1 per write, on each rank. This lets us use it
// For performance, we append <data_ptr> directly into the linked list // to compute a global part-numbering.
// of data in param->io_buf. We are "appending" rather than //
// "extending", so the added buffer is seen as written data, rather // In the N:N case, we only need to increment part-numbers within
// than empty storage. // each rank.
aws_iobuf_reset(param->io_buf); //
aws_iobuf_append_static(param->io_buf, data_ptr, remaining); // In the N:1 case, the global order of part-numbers we're writing
AWS4C_CHECK( s3_put(param->io_buf, file) ); // depends on whether wer're writing strided or segmented, in
AWS4C_CHECK_OK( param->io_buf ); // other words, how <offset> and <remaining> are acutally
// positioning the parts being written. [See discussion at
// drop ptrs to <data_ptr>, in param->io_buf // S3_Close_internal().]
aws_iobuf_reset(param->io_buf); //
}
else if (multi_part_upload_p) { // N:1 (with MPU)
// Ordering of the part-numbers imposes a global ordering on
// the components of the final object. param->part_number
// is incremented by 1 per write, on each rank. This lets us
// use it to compute a global part-numbering.
//
// NOTE: 's3curl.pl --debug' shows StringToSign having partNumber // NOTE: 's3curl.pl --debug' shows StringToSign having partNumber
// first, even if I put uploadId first in the URL. Maybe // first, even if I put uploadId first in the URL. Maybe
// that's what the server will do. GetStringToSign() in // that's what the server will do. GetStringToSign() in
// aws4c is not clever about this, so we spoon-feed args in // aws4c is not clever about this, so we spoon-feed args in
// the proper order. // the proper order.
size_t part_number = (param->part_number++ * numTasksWorld) + rank; size_t part_number;
if (n_to_1) {
if (segmented) { // segmented
size_t parts_per_rank = param->blockSize / param->transferSize;
part_number = (rank * parts_per_rank) + param->part_number;
}
else // strided
part_number = (param->part_number * param->numTasks) + rank;
}
else
part_number = param->part_number;
++ param->part_number;
// if (verbose >= VERBOSE_3) {
// fprintf( stdout, "rank %d of %d writing (%s,%s) part_number %lld\n",
// rank,
// param->numTasks,
// (n_to_1 ? "N:1" : "N:N"),
// (segmented ? "segmented" : "strided"),
// part_number);
// }
snprintf(buff, BUFF_SIZE, snprintf(buff, BUFF_SIZE,
"%s?partNumber=%d&uploadId=%s", "%s?partNumber=%d&uploadId=%s",
fname, part_number, param->UploadId); fname, part_number, param->UploadId);
@ -722,14 +735,30 @@ S3_Xfer_internal(int access,
AWS4C_CHECK( s3_put(param->io_buf, buff) ); AWS4C_CHECK( s3_put(param->io_buf, buff) );
AWS4C_CHECK_OK( param->io_buf ); AWS4C_CHECK_OK( param->io_buf );
if (verbose >= VERBOSE_4) { // if (verbose >= VERBOSE_3) {
printf("rank %d: read ETag = '%s'\n", rank, param->io_buf->eTag); // printf("rank %d: read ETag = '%s'\n", rank, param->io_buf->eTag);
if (strlen(param->io_buf->eTag) != ETAG_SIZE+2) { /* quotes at both ends */ // if (strlen(param->io_buf->eTag) != ETAG_SIZE+2) { /* quotes at both ends */
// fprintf(stderr, "Rank %d: ERROR: expected ETag to be %d hex digits\n",
// rank, ETAG_SIZE);
// exit(1);
// }
// }
if (verbose >= VERBOSE_3) {
fprintf( stdout, "rank %d of %d (%s,%s) offset %lld, part# %lld --> ETag %s\n",
rank,
param->numTasks,
(n_to_1 ? "N:1" : "N:N"),
(segmented ? "segmented" : "strided"),
offset,
part_number,
param->io_buf->eTag); // incl quote-marks at [0] and [len-1]
}
if (strlen(param->io_buf->eTag) != ETAG_SIZE+2) { /* quotes at both ends */
fprintf(stderr, "Rank %d: ERROR: expected ETag to be %d hex digits\n", fprintf(stderr, "Rank %d: ERROR: expected ETag to be %d hex digits\n",
rank, ETAG_SIZE); rank, ETAG_SIZE);
exit(1); exit(1);
} }
}
// save the eTag for later // save the eTag for later
// //
@ -746,10 +775,15 @@ S3_Xfer_internal(int access,
// drop ptrs to <data_ptr>, in param->io_buf // drop ptrs to <data_ptr>, in param->io_buf
aws_iobuf_reset(param->io_buf); aws_iobuf_reset(param->io_buf);
} }
else { // N:1 (use EMC's byte-range write support, instead of MPU) else { // use EMC's byte-range write-support, instead of MPU
// NOTE: You must call 's3_enable_EMC_extensions(1)' to let this work:
s3_set_byte_range(offset, remaining); // produces appropriate byte-range header // NOTE: You must call 's3_enable_EMC_extensions(1)' for
// byte-ranges to work for writes.
if (n_to_n)
s3_set_byte_range(-1,-1); // EMC header "Range: bytes=-1-" means "append"
else
s3_set_byte_range(offset, remaining);
// For performance, we append <data_ptr> directly into the linked list // For performance, we append <data_ptr> directly into the linked list
// of data in param->io_buf. We are "appending" rather than // of data in param->io_buf. We are "appending" rather than
@ -757,7 +791,7 @@ S3_Xfer_internal(int access,
// than empty storage. // than empty storage.
aws_iobuf_reset(param->io_buf); aws_iobuf_reset(param->io_buf);
aws_iobuf_append_static(param->io_buf, data_ptr, remaining); aws_iobuf_append_static(param->io_buf, data_ptr, remaining);
AWS4C_CHECK( s3_put(param->io_buf, file) ); AWS4C_CHECK ( s3_put(param->io_buf, file) );
AWS4C_CHECK_OK( param->io_buf ); AWS4C_CHECK_OK( param->io_buf );
// drop ptrs to <data_ptr>, in param->io_buf // drop ptrs to <data_ptr>, in param->io_buf
@ -772,13 +806,14 @@ S3_Xfer_internal(int access,
} }
else { /* READ or CHECK */ else { /* READ or CHECK */
if (verbose >= VERBOSE_4) { if (verbose >= VERBOSE_3) {
fprintf( stdout, "task %d reading from offset %lld\n", fprintf( stdout, "rank %d reading from offset %lld\n",
rank, rank,
param->offset + length - remaining ); param->offset + length - remaining );
} }
// read specific byte-range from the object // read specific byte-range from the object
// [This is included in the "pure" S3 spec.]
s3_set_byte_range(offset, remaining); s3_set_byte_range(offset, remaining);
// For performance, we append <data_ptr> directly into the linked // For performance, we append <data_ptr> directly into the linked
@ -792,12 +827,12 @@ S3_Xfer_internal(int access,
AWS4C_CHECK( s3_get(param->io_buf, file) ); AWS4C_CHECK( s3_get(param->io_buf, file) );
if (param->io_buf->code != 206) { /* '206 Partial Content' */ if (param->io_buf->code != 206) { /* '206 Partial Content' */
snprintf(buff, BUFF_SIZE, snprintf(buff, BUFF_SIZE,
"ERROR: Unexpected result (%d, '%s') at %s, line %d\n", "Unexpected result (%d, '%s')",
param->io_buf->code, param->io_buf->result, __FILE__, __LINE__); param->io_buf->code, param->io_buf->result);
ERR_SIMPLE(buff); ERR_SIMPLE(buff);
} }
// drop ptrs to <data_ptr>, in param->io_buf // drop refs to <data_ptr>, in param->io_buf
aws_iobuf_reset(param->io_buf); aws_iobuf_reset(param->io_buf);
} }
@ -908,173 +943,282 @@ S3_Close_internal( void* fd,
IOR_param_t* param, IOR_param_t* param,
int multi_part_upload_p ) { int multi_part_upload_p ) {
if (param->verbose >= VERBOSE_2) {
printf("-> S3_Close\n");
}
char* fname = (char*)fd; /* see NOTE above S3_Create_Or_Open() */ char* fname = (char*)fd; /* see NOTE above S3_Create_Or_Open() */
// easier to think
int n_to_n = param->filePerProc;
int n_to_1 = (! n_to_n);
int segmented = (param->segmentCount == 1);
if (param->verbose >= VERBOSE_2) {
printf("-> S3_Close('%s', ,%d) %s\n",
fname,
multi_part_upload_p,
((n_to_n) ? "N:N" : ((segmented) ? "N:1(seg)" : "N:1(str)")));
}
if (param->open == WRITE) { if (param->open == WRITE) {
// closing N:N write
if (param->filePerProc) {
// nothing to do ...
}
// closing N:1 write (with Multi-Part Upload) // finalizing Multi-Part Upload (for N:1 or N:N)
else if (multi_part_upload_p) { if (multi_part_upload_p) {
MPI_Datatype mpi_size_t;
if (sizeof(size_t) == sizeof(int))
mpi_size_t = MPI_INT;
else if (sizeof(size_t) == sizeof(long))
mpi_size_t = MPI_LONG;
else
mpi_size_t = MPI_LONG_LONG;
// Everybody should have the same number of ETags (?) size_t etag_data_size = param->etags->write_count; /* local ETag data (bytes) */
size_t etag_data_size = param->etags->write_count; /* size of local ETag data */ size_t etags_per_rank = etag_data_size / ETAG_SIZE; /* number of local etags */
size_t etag_count = etag_data_size / ETAG_SIZE; /* number of local etags */
size_t etag_count_max = 0; /* highest number on any proc */
MPI_Allreduce(&etag_count, &etag_count_max, // --- create XML containing ETags in an IOBuf for "close" request
1, mpi_size_t, MPI_MAX, param->testComm); IOBuf* xml = NULL;
if (etag_count != etag_count_max) {
printf("Rank %d: etag count mismatch: max:%d, mine:%d\n",
rank, etag_count_max, etag_count);
MPI_Abort(param->testComm, 1);
}
// collect ETag data at Rank0
aws_iobuf_realloc(param->etags); /* force single contiguous buffer */
char* etag_data = param->etags->first->buf; /* ptr to contiguous data */
if (rank != 0) { if (n_to_1) {
MPI_Gather(etag_data, etag_data_size, MPI_BYTE,
NULL, etag_data_size, MPI_BYTE, 0, MPI_COMM_WORLD);
}
else {
char* etag_ptr;
int rnk;
int i;
char* etag_vec = (char*)malloc((numTasksWorld * etag_data_size) +1); // for N:1, gather all Etags at Rank0
if (! etag_vec) { MPI_Datatype mpi_size_t;
fprintf(stderr, "rank 0 failed to malloc %d bytes\n", if (sizeof(size_t) == sizeof(int))
numTasksWorld * etag_data_size); mpi_size_t = MPI_INT;
else if (sizeof(size_t) == sizeof(long))
mpi_size_t = MPI_LONG;
else
mpi_size_t = MPI_LONG_LONG;
// Everybody should have the same number of ETags (?)
size_t etag_count_max = 0; /* highest number on any proc */
MPI_Allreduce(&etags_per_rank, &etag_count_max,
1, mpi_size_t, MPI_MAX, param->testComm);
if (etags_per_rank != etag_count_max) {
printf("Rank %d: etag count mismatch: max:%d, mine:%d\n",
rank, etag_count_max, etags_per_rank);
MPI_Abort(param->testComm, 1); MPI_Abort(param->testComm, 1);
} }
MPI_Gather(etag_data, etag_data_size, MPI_BYTE,
etag_vec, etag_data_size, MPI_BYTE, 0, MPI_COMM_WORLD);
// --- debugging: show the gathered etag data // collect ETag data at Rank0
// (This shows the raw concatenated etag-data from each node.) aws_iobuf_realloc(param->etags); /* force single contiguous buffer */
if (param->verbose >= VERBOSE_4) { char* etag_data = param->etags->first->buf; /* per-rank data, contiguous */
printf("rank 0: gathered %d etags from all ranks:\n", etag_count);
etag_ptr=etag_vec;
for (rnk=0; rnk<numTasksWorld; ++rnk) {
printf("\t[%d]: '", rnk);
int ii; if (rank == 0) {
for (ii=0; ii<etag_data_size; ++ii) /* NOT null-terminated! */ char* etag_ptr;
printf("%c", etag_ptr[ii]); int i;
int j;
int rnk;
printf("'\n"); char* etag_vec = (char*)malloc((param->numTasks * etag_data_size) +1);
etag_ptr += etag_data_size; if (! etag_vec) {
fprintf(stderr, "rank 0 failed to malloc %d bytes\n",
param->numTasks * etag_data_size);
MPI_Abort(param->testComm, 1);
} }
MPI_Gather(etag_data, etag_data_size, MPI_BYTE,
etag_vec, etag_data_size, MPI_BYTE, 0, MPI_COMM_WORLD);
// --- debugging: show the gathered etag data
// (This shows the raw concatenated etag-data from each node.)
if (param->verbose >= VERBOSE_4) {
printf("rank 0: gathered %d etags from all ranks:\n", etags_per_rank);
etag_ptr=etag_vec;
for (rnk=0; rnk<param->numTasks; ++rnk) {
printf("\t[%d]: '", rnk);
int ii;
for (ii=0; ii<etag_data_size; ++ii) /* NOT null-terminated! */
printf("%c", etag_ptr[ii]);
printf("'\n");
etag_ptr += etag_data_size;
}
}
// add XML for *all* the parts. The XML must be ordered by
// part-number. Each rank wrote <etags_per_rank> parts,
// locally. At rank0, the etags for each rank are now
// stored as a continguous block of text, with the blocks
// stored in rank order in etag_vec. In other words, our
// internal rep at rank 0 matches the "segmented" format.
// From this, we must select etags in an order matching how
// they appear in the actual object, and give sequential
// part-numbers to the resulting sequence.
//
// That ordering of parts in the actual written object
// varies according to whether we wrote in the "segmented"
// or "strided" format.
//
// supposing N ranks, and P parts per rank:
//
// segmented:
//
// all parts for a given rank are consecutive.
// rank r writes these parts:
//
// rP, rP+1, ... (r+1)P -1
//
// i.e. rank0 writes parts 0,1,2,3 ... P-1
//
//
// strided:
//
// rank r writes every P-th part, starting with r.
//
// r, P+r, ... (P-1)P + r
//
// i.e. rank0 writes parts 0,P,2P,3P ... (P-1)P
//
//
// NOTE: If we knew ahead of time how many parts each rank was
// going to write, we could assign part-number ranges, per
// rank, and then have nice locality here.
//
// Alternatively, we could have everyone format their own
// XML text and send that, instead of just the tags. This
// would increase the amount of data being sent, but would
// reduce the work for rank0 to format everything.
size_t i_max; // outer-loop
size_t j_max; // inner loop
size_t start_multiplier; // initial offset in collected data
size_t stride; // in etag_vec
if (segmented) { // segmented
i_max = param->numTasks;
j_max = etags_per_rank;
start_multiplier = etag_data_size; /* one rank's-worth of Etag data */
stride = ETAG_SIZE; /* one ETag */
}
else { // strided
i_max = etags_per_rank;
j_max = param->numTasks;
start_multiplier = ETAG_SIZE; /* one ETag */
stride = etag_data_size; /* one rank's-worth of Etag data */
}
xml = aws_iobuf_new();
aws_iobuf_growth_size(xml, 1024 * 8);
// write XML header ...
aws_iobuf_append_str(xml, "<CompleteMultipartUpload>\n");
int part = 0;
for (i=0; i<i_max; ++i) {
etag_ptr=etag_vec + (i * start_multiplier);
for (j=0; j<j_max; ++j) {
// etags were saved as contiguous text. Extract the next one.
char etag[ETAG_SIZE +1];
memcpy(etag, etag_ptr, ETAG_SIZE);
etag[ETAG_SIZE] = 0;
// write XML for next part, with Etag ...
snprintf(buff, BUFF_SIZE,
" <Part>\n"
" <PartNumber>%d</PartNumber>\n"
" <ETag>%s</ETag>\n"
" </Part>\n",
part, etag);
aws_iobuf_append_str(xml, buff);
etag_ptr += stride;
++ part;
}
}
// write XML tail ...
aws_iobuf_append_str(xml, "</CompleteMultipartUpload>\n");
} }
// --- create XML containing ETags in an IOBuf for "close" request else {
IOBuf* xml = aws_iobuf_new(); MPI_Gather(etag_data, etag_data_size, MPI_BYTE,
NULL, etag_data_size, MPI_BYTE, 0, MPI_COMM_WORLD);
}
}
else { /* N:N */
xml = aws_iobuf_new();
aws_iobuf_growth_size(xml, 1024 * 8); aws_iobuf_growth_size(xml, 1024 * 8);
// write XML header ... // write XML header ...
aws_iobuf_append_str(xml, "<CompleteMultipartUpload>\n"); aws_iobuf_append_str(xml, "<CompleteMultipartUpload>\n");
// add XML for *all* the parts. The XML must be ordered by // all parts of our object were written from this rank.
// part-number. Each rank wrote <etag_count> parts. The etags char etag[ETAG_SIZE +1];
// for each rank are staored as a continguous block of text, with int part = 0;
// the blocks stored in rank order in etag_vec. We must therefore int i;
// access them in the worst possible way, regarding locality. for (i=0; i<etags_per_rank; ++i) {
//
// NOTE: If we knew ahead of time how many parts each rank was
// going to write, we could assign part-number ranges, per
// rank, and then have nice locality here.
//
// Alternatively, we could have everyone format their own
// XML text and send that, instead of just the tags. This
// would increase the amount of data being sent, but would
// reduce the work for rank0 to format everything.
int part = 0;
for (i=0; i<etag_count; ++i) {
etag_ptr=etag_vec + (i * ETAG_SIZE);
for (rnk=0; rnk<numTasksWorld; ++rnk) { // TBD: Instead of reading into etag, then sprintf'ing, then
// copying into xml, we could just read directly into xml
// etags were saved as contiguous text. Extract the next one. int sz = aws_iobuf_get_raw(param->etags, etag, ETAG_SIZE);
char etag[ETAG_SIZE +1]; if (sz != ETAG_SIZE) {
memcpy(etag, etag_ptr, ETAG_SIZE);
etag[ETAG_SIZE] = 0;
// write XML for next part, with Etag ...
snprintf(buff, BUFF_SIZE, snprintf(buff, BUFF_SIZE,
" <Part>\n" "Read of ETag %d had length %d (not %d)\n",
" <PartNumber>%d</PartNumber>\n" rank, i, sz, ETAG_SIZE);
" <ETag>%s</ETag>\n" ERR_SIMPLE(buff);
" </Part>\n",
part, etag);
aws_iobuf_append_str(xml, buff);
etag_ptr += etag_data_size;
++ part;
} }
etag[ETAG_SIZE] = 0;
// write XML for next part, with Etag ...
snprintf(buff, BUFF_SIZE,
" <Part>\n"
" <PartNumber>%d</PartNumber>\n"
" <ETag>%s</ETag>\n"
" </Part>\n",
part, etag);
aws_iobuf_append_str(xml, buff);
++ part;
} }
// write XML tail ... // write XML tail ...
aws_iobuf_append_str(xml, "</CompleteMultipartUpload>\n"); aws_iobuf_append_str(xml, "</CompleteMultipartUpload>\n");
}
// send request to finalize MPU
if (n_to_n || (rank == 0)) {
// DEBUGGING: show the XML we constructed // DEBUGGING: show the XML we constructed
if (param->verbose >= VERBOSE_4) if (param->verbose >= VERBOSE_3)
debug_iobuf(xml, 1, 1); debug_iobuf(xml, 1, 1);
// --- POST our XML to the server. // --- POST our XML to the server.
snprintf(buff, BUFF_SIZE, snprintf(buff, BUFF_SIZE,
"%s?uploadId=%s", "%s?uploadId=%s",
fname, param->UploadId); fname, param->UploadId);
#if 1
AWS4C_CHECK ( s3_post(xml, buff) ); AWS4C_CHECK ( s3_post(xml, buff) );
AWS4C_CHECK_OK( xml ); AWS4C_CHECK_OK( xml );
#else
IOBuf* response = aws_iobuf_new();
aws_iobuf_reset(response);
AWS4C_CHECK( s3_post2(xml, buff, NULL, response) );
if (! AWS4C_OK(param->io_buf) ) {
fprintf(stderr, "rank %d: POST '%s' failed: %s\n",
rank, buff, param->io_buf->result);
int sz;
for (sz = aws_iobuf_getline(response, buff, BUFF_SIZE);
sz;
sz = aws_iobuf_getline(response, buff, BUFF_SIZE)) {
printf("-- %s\n", buff);
}
MPI_Abort(param->testComm, 1);
}
aws_iobuf_free(response);
#endif
aws_iobuf_free(xml); aws_iobuf_free(xml);
} }
// --- reset associated info. allows another MPU, and frees memory. // everybody reset MPU info. Allows another MPU, and frees memory.
s3_MPU_reset(param); s3_MPU_reset(param);
// Don't you non-zero ranks go trying to stat the N:1 file until // Everybody meetup, so non-zero ranks won't go trying to stat the
// rank0 has finished the S3 multi-part finalize. It will not appear // N:1 file until rank0 has finished the S3 multi-part finalize.
// to exist, until then. // The object will not appear to exist, until then.
MPI_CHECK(MPI_Barrier(param->testComm), "barrier error"); if (n_to_1)
MPI_CHECK(MPI_Barrier(param->testComm), "barrier error");
}
else {
// No finalization is needed, when using EMC's byte-range writing
// support. However, we do need to make sure everyone has
// finished writing, before anyone starts reading.
if (n_to_1) {
MPI_CHECK(MPI_Barrier(param->testComm), "barrier error");
if (param->verbose >= VERBOSE_2)
printf("rank %d: passed barrier\n", rank);
}
} }
// After writing, reset the CURL connection, so that caches won't be // After writing, reset the CURL connection, so that caches won't be
@ -1123,7 +1267,17 @@ S3_Delete( char *testFileName, IOR_param_t * param ) {
/* maybe initialize curl */ /* maybe initialize curl */
s3_connect( param ); s3_connect( param );
#if 0
// EMC BUG: If file was written with appends, and is deleted,
// Then any future recreation will result in an object that can't be read.
// this
AWS4C_CHECK( s3_delete(param->io_buf, testFileName) ); AWS4C_CHECK( s3_delete(param->io_buf, testFileName) );
#else
// just replace with a zero-length object for now
aws_iobuf_reset(param->io_buf);
AWS4C_CHECK ( s3_put(param->io_buf, testFileName) );
#endif
AWS4C_CHECK_OK( param->io_buf ); AWS4C_CHECK_OK( param->io_buf );
if (param->verbose >= VERBOSE_2) if (param->verbose >= VERBOSE_2)
@ -1218,6 +1372,9 @@ S3_GetFileSize(IOR_param_t * param,
} }
aggFileSizeFromStat = param->io_buf->contentLen; aggFileSizeFromStat = param->io_buf->contentLen;
if (param->verbose >= VERBOSE_2) {
printf("\trank %d: file-size %llu\n", rank, aggFileSizeFromStat);
}
if ( param->filePerProc == TRUE ) { if ( param->filePerProc == TRUE ) {
if (param->verbose >= VERBOSE_2) { if (param->verbose >= VERBOSE_2) {

6
src/ior.c
View File

@ -263,9 +263,7 @@ static void AioriBind(char* api, IOR_param_t* param)
ERR("unrecognized IO API"); ERR("unrecognized IO API");
} }
else if (! strncmp(api, "S3", 2)) { else if (! strncmp(api, "S3", 2)) {
if (! strcmp(api, "S3_plus")) if (! strcmp(api, "S3_EMC"))
param->curl_flags |= IOR_CURL_S3_EMC_EXT;
else if (! strcmp(api, "S3_EMC"))
param->curl_flags |= IOR_CURL_S3_EMC_EXT; param->curl_flags |= IOR_CURL_S3_EMC_EXT;
else else
param->curl_flags &= ~(IOR_CURL_S3_EMC_EXT); param->curl_flags &= ~(IOR_CURL_S3_EMC_EXT);
@ -742,7 +740,7 @@ static void DisplayUsage(char **argv)
{ {
char *opts[] = { char *opts[] = {
"OPTIONS:", "OPTIONS:",
" -a S api -- API for I/O [POSIX|MPIIO|HDF5|NCMPI]", " -a S api -- API for I/O [POSIX|MPIIO|HDF5|HDFS|S3|S3_EMC|NCMPI]",
" -A N refNum -- user supplied reference number to include in the summary", " -A N refNum -- user supplied reference number to include in the summary",
" -b N blockSize -- contiguous bytes to write per task (e.g.: 8, 4k, 2m, 1g)", " -b N blockSize -- contiguous bytes to write per task (e.g.: 8, 4k, 2m, 1g)",
" -B useO_DIRECT -- uses O_DIRECT for POSIX, bypassing I/O buffers", " -B useO_DIRECT -- uses O_DIRECT for POSIX, bypassing I/O buffers",