March 2012 - Triton-commits - mailman3.cels.anl.gov

Triton Repository branch, master, updated. fc5bb723bc1698c1ad5390453c53ad0f2e6c7190
by noreply＠mcs.anl.gov 20 Mar '12

20 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton Repository". The branch, master has been updated via fc5bb723bc1698c1ad5390453c53ad0f2e6c7190 (commit) from 2595f40481c999ba6f5e184d85f788fa09711dd5 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit fc5bb723bc1698c1ad5390453c53ad0f2e6c7190 Author: Phil Carns <carns(a)mcs.anl.gov> Date: Tue Mar 20 15:24:57 2012 -0400 bug fixes: basic test working now ----------------------------------------------------------------------- Summary of changes: code/src/transactional-osd/tests/tosd-bench-idb.ae | 5 +---- 1 files changed, 1 insertions(+), 4 deletions(-) Diff of changes: diff --git a/code/src/transactional-osd/tests/tosd-bench-idb.ae b/code/src/transactional-osd/tests/tosd-bench-idb.ae index 9fb59f4..1789112 100644 --- a/code/src/transactional-osd/tests/tosd-bench-idb.ae +++ b/code/src/transactional-osd/tests/tosd-bench-idb.ae @@ -197,7 +197,7 @@ __blocking int aesop_main(int argc, char **argv) #else ret = sscanf(line, "%lu %s %ld %ld", &tmp_op->oid.l, op_string, &tmp_op->offset, &tmp_op->size); #endif - if(ret != 3) + if(ret != 4) { fprintf(stderr, "Error: bad line: %s\n", line); return(-1); @@ -248,9 +248,6 @@ static triton_ret_t init_tosd_zeroconf_params(int argc, char** argv) tret = triton_zeroconf_set("triton.tosd.data_path", tosd_path_logs); triton_error_assert(tret); - tret = triton_zeroconf_set("triton.tosd.alignment", "1"); - triton_error_assert(tret); - return(TRITON_SUCCESS); } hooks/post-receive -- Triton Repository

1 0

Triton Repository branch, master, updated. 2595f40481c999ba6f5e184d85f788fa09711dd5
by noreply＠mcs.anl.gov 20 Mar '12

20 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton Repository". The branch, master has been updated via 2595f40481c999ba6f5e184d85f788fa09711dd5 (commit) from db91cbfc9274581d604c729f720963192b3519f0 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 2595f40481c999ba6f5e184d85f788fa09711dd5 Author: Phil Carns <carns(a)mcs.anl.gov> Date: Tue Mar 20 15:18:34 2012 -0400 update object rebuild doc - reflect discussion from earlier today: - pull rebuild data from a random replica - include rationale for that design choice - reference section elsewhere in doc for optimization strategies - update outline of rebuild procedure to refer to random replicas rather than "parent" replicas in the chain ----------------------------------------------------------------------- Summary of changes: code/doc/resilience/object-rebuild.txt | 34 +++++++++++++++++++++---------- 1 files changed, 23 insertions(+), 11 deletions(-) Diff of changes: diff --git a/code/doc/resilience/object-rebuild.txt b/code/doc/resilience/object-rebuild.txt index 85afba6..74ea09e 100644 --- a/code/doc/resilience/object-rebuild.txt +++ b/code/doc/resilience/object-rebuild.txt @@ -255,8 +255,8 @@ reconstruction. [[rebuild]] The Triton rebuild process is driven by servers that hold secondary objects -that are out of sync with their primaries (i.e., stubs). These servers "pull" data -from primary (or parent) objects. Once they have pulled all necessary data +that are out of sync with their primaries (i.e., stubs). These servers +"pull" data from complete objects in the replica set. Once they have pulled all necessary data for a given object, then the stub designation will be removed from that object and clients will be allowed to interact with it as usual. @@ -270,8 +270,9 @@ optimal order to perform rebuilds. already- see the replication documentation for details) ... may be "0" if it this is a new stub with no data -.. contact parent to request rebuild data -.. on parent: +.. select a random peer server in the replica set +.. contact the random peer to request data +.. on peer: ... checkpoint the object in question ... stream a _pickled_ (i.e,, serialized) copy of any object data that is newer than the replica's update state to the replica @@ -284,17 +285,28 @@ The objects that have partial data should be recovered first. Those objects represent replicas that are mostly complete but just lack updates that were in flight at the time of failure. +Rationale +[NOTE] +When a server is trying to pull rebuild data for a given object, it can pull +data from any existing replica. This could be a server adjacent to the +rebuiling server, a server adjacent to the the failure, or any server +holding a replica for an object in question. In the prototype phase, we +will select a replica at random in an attempt to distribute load. In the +longer run we could optimize this either locally (by attempting to +distribute pull requests evenly) or with a global optimization. More +information can be found in a later section of this document. + ==== Additional failures during rebuild -* If parent detects a child failure while transmitting serialized data: +* If failure detected while transmitting pickled data: ** do nothing except terminate data transfer and remove snapshot ** subsequent recovery and rebuild steps will restart the transfer if needed -* If a child replica attempts to rebuild from a parent, but that parent also -turns out to be a stub: -** attempt rebuild from parent of parent +* If the server contacted to pull data from also holds a stub for the object: +** attempt rebuild from a server closer to the OID in the placement +algorithm -* If child detects that all parents are stubs: +* If child detects that all replicas are stubs: ** keep retrying until primary figures out whats going on (see below) * If a primary is a stub (there are no parents to rebuild from): @@ -309,8 +321,8 @@ were in flight, but at least one replica survived: to multiple failures: **** Data loss, notify administrator -* If a child detects a parent failure while receiving serialized data: -** restart rebuild to next parent +* If a stub detects a failure while receiving pickled data: +** restart rebuild pulling from another random server in the replica set ==== Scheduling data transfers hooks/post-receive -- Triton Repository

1 0

Triton Repository branch, master, updated. db91cbfc9274581d604c729f720963192b3519f0
by noreply＠mcs.anl.gov 20 Mar '12

20 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton Repository". The branch, master has been updated via db91cbfc9274581d604c729f720963192b3519f0 (commit) via 7eed0320a6b7d0510861c3806e11e4ef382fba51 (commit) from 30af1a8132ee1a74c890e9a4d7d2302e3715e14d (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit db91cbfc9274581d604c729f720963192b3519f0 Author: Phil Carns <carns(a)mcs.anl.gov> Date: Tue Mar 20 15:08:13 2012 -0400 incremental progress commit 7eed0320a6b7d0510861c3806e11e4ef382fba51 Author: Phil Carns <carns(a)mcs.anl.gov> Date: Tue Mar 20 12:52:00 2012 -0400 strip out some cruft ----------------------------------------------------------------------- Summary of changes: code/src/transactional-osd/tests/tosd-bench-idb.ae | 139 ++++++++++++-------- 1 files changed, 82 insertions(+), 57 deletions(-) Diff of changes: diff --git a/code/src/transactional-osd/tests/tosd-bench-idb.ae b/code/src/transactional-osd/tests/tosd-bench-idb.ae index 579cdcb..9fb59f4 100644 --- a/code/src/transactional-osd/tests/tosd-bench-idb.ae +++ b/code/src/transactional-osd/tests/tosd-bench-idb.ae @@ -14,109 +14,130 @@ enum op_type READ }; +/* struct describing each operation to perform as specified in the input + * workload file + */ struct bench_op { + uint128_t oid; int64_t offset; int64_t size; enum op_type type; struct triton_list_link list_link; }; +/* list of operations to perform */ TRITON_LIST_DEFINE(op_list); triton_mutex_t op_list_mutex = TRITON_MUTEX_INITIALIZER; +/* timing routine */ static double Wtime(void) { struct timeval t; gettimeofday(&t, NULL); return((double)t.tv_sec + (double)(t.tv_usec) / 1000000); } + static double start_tm; static double end_tm; static int concurrent; -static uint64_t first_version; -static uint64_t version; -static int max_seconds = 0; -static uint128_t oid; +static int numoids = 0; +static int total_ops = 0; +static int64_t max_size = 0; static __blocking int do_tosd_test(void) { triton_ret_t tret; - int w_flags = TOSD_FLAG_AUTO_TXN; - - tret = tosd_create(oid, 0); - if(tret == TRITON_ERR_EXIST) + int i; + uint64_t* version_array; + char** buffer_array; + uint128_t oid; + + /* allocate a buffer big enough to hold the biggest + * operation in the workload file + */ + buffer_array = malloc(concurrent * sizeof(*buffer_array)); + assert(buffer_array); + for(i=0; i<concurrent; i++) { - printf("# operating on existing object.\n"); - triton_error_destroy(tret); - tret = TRITON_SUCCESS; + buffer_array[i] = malloc(max_size); + assert(buffer_array[i]); } - triton_error_assert(tret); - tret = tosd_get_version(oid, &version); - triton_error_assert(tret); + /* array to track current version for each object */ + version_array = malloc((numoids+1)*sizeof(*version_array)); + assert(version_array); + + /* create each object (if it doesn't exist) and retrieve its current + * version + */ + for(i=1; i<= numoids; i++) + { + oid.u = 0; + oid.l = i; + tret = tosd_create(oid, 0); + if(tret == TRITON_ERR_EXIST) + { + printf("# operating on existing object (%d).\n", i); + triton_error_destroy(tret); + tret = TRITON_SUCCESS; + } + triton_error_assert(tret); - first_version = version; + tret = tosd_get_version(oid, &version_array[i]); + triton_error_assert(tret); + } + /* start timing */ start_tm = Wtime(); - pwait { pprivate int i; pprivate struct triton_list_link* tmp_link; - pprivate char* buffer = NULL; - pprivate uint64_t b_version; + pprivate uint64_t this_version; pprivate struct bench_op* tmp_op; pprivate triton_ret_t b_ret; pprivate int64_t out_size; - pprivate int64_t keyval_offset = 0; - pprivate int64_t* offset_ptr; - pprivate uint64_t fork; - pprivate int ret; - pprivate double now; + /* separate pbranch for each concurrent operator */ for(i=0; i<concurrent; i++) { pbranch { triton_mutex_lock(&op_list_mutex); + /* get the next operation off of the list */ tmp_link = triton_stack_pop(&op_list); - triton_mutex_unlock(&op_list_mutex); - - while(tmp_link) + if(tmp_link) { - b_version = version++; - tmp_op = triton_list_get_entry(tmp_link, struct bench_op, list_link); + /* increment version number for this object */ + this_version = ++version_array[tmp_op->oid.l]; + } + triton_mutex_unlock(&op_list_mutex); - offset_ptr = &tmp_op->offset; - fork = 0; - + while(tmp_link) + { if(tmp_op->type == WRITE) { - b_ret = tosd_write(oid, fork, b_version, &buffer, + b_ret = tosd_write(tmp_op->oid, 0, this_version, &buffer_array[i], &tmp_op->size, 1, &tmp_op->offset, - &tmp_op->size, 1, w_flags, 0); + &tmp_op->size, 1, TOSD_FLAG_AUTO_TXN, 0); triton_error_assert(b_ret); } else { assert(tmp_op->type == READ); - b_ret = tosd_read(oid, fork, &buffer, &tmp_op->size, 1, + b_ret = tosd_read(tmp_op->oid, 0, &buffer_array[i], &tmp_op->size, 1, &tmp_op->offset, &tmp_op->size, 1, &out_size, 0); triton_error_assert(b_ret); assert(out_size == tmp_op->size); } - now = Wtime(); triton_mutex_lock(&op_list_mutex); - /* check the time limit if applicable */ - if(max_seconds > 0 && (now - start_tm) > max_seconds) - tmp_link = NULL; - else - tmp_link = triton_stack_pop(&op_list); + total_ops++; + tmp_link = triton_stack_pop(&op_list); triton_mutex_unlock(&op_list_mutex); } } @@ -130,23 +151,13 @@ static __blocking int do_tosd_test(void) __blocking int aesop_main(int argc, char **argv) { - ae_context_t ctx; - ae_hints_t hints; - ae_op_id_t op_id; - int pc = 0; FILE *desc = 0; char line[2048]; char op_string[100]; - int64_t size; - int64_t offset; int ret; struct bench_op* tmp_op; - triton_ret_t tret; - const char* tosd_init_flags = NULL; int i; - oid = triton_uint128_from_uint64(1); - if(argc != 5) { fprintf(stderr, "Usage: tosd-bench-concurrent <workload description file> <dir> <concurrency> <threaded y|n>\n"); @@ -171,18 +182,31 @@ __blocking int aesop_main(int argc, char **argv) { if(line[0] == '#') continue; + if((ret = sscanf(line, "oids=%d", &numoids)) == 1) + { + /* parsed line containing num oids */ + continue; + } + + tmp_op = malloc(sizeof(*tmp_op)); + assert(tmp_op); + memset(tmp_op, 0, sizeof(*tmp_op)); + #if SIZEOF_LONG_INT == 4 - ret = sscanf(line, "%s %lld %lld", op_string, &offset, &size); + ret = sscanf(line, "%llu %s %lld %lld", &tmp_op->oid.l, op_string, &tmp_op->offset, &tmp_op->size); #else - ret = sscanf(line, "%s %ld %ld", op_string, &offset, &size); + ret = sscanf(line, "%lu %s %ld %ld", &tmp_op->oid.l, op_string, &tmp_op->offset, &tmp_op->size); #endif if(ret != 3) { fprintf(stderr, "Error: bad line: %s\n", line); return(-1); } - tmp_op = malloc(sizeof(*tmp_op)); - assert(tmp_op); + assert(tmp_op->oid.l <= numoids); + + /* track size of largest operation in the workload */ + if(tmp_op->size > max_size) + max_size = tmp_op->size; if(strcmp(op_string, "write") == 0) tmp_op->type = WRITE; @@ -190,12 +214,13 @@ __blocking int aesop_main(int argc, char **argv) tmp_op->type = READ; else assert(0); - tmp_op->offset = offset; - tmp_op->size = size; + triton_list_add_back(&tmp_op->list_link, &op_list); } fclose(desc); + assert(numoids > 0); + do_tosd_test(); /* print command line in comments */ @@ -206,7 +231,7 @@ __blocking int aesop_main(int argc, char **argv) } printf("\n"); printf("# <seconds> <ops/s>\n"); - printf("%f\t%f\n", end_tm-start_tm, ((double)(version-first_version))/(end_tm-start_tm)); + printf("%f\t%f\n", end_tm-start_tm, ((double)(total_ops))/(end_tm-start_tm)); return 0; } hooks/post-receive -- Triton Repository

1 0

Triton Repository branch, master, updated. 30af1a8132ee1a74c890e9a4d7d2302e3715e14d
by noreply＠mcs.anl.gov 20 Mar '12

20 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton Repository". The branch, master has been updated via 30af1a8132ee1a74c890e9a4d7d2302e3715e14d (commit) via b88b17823f9e28705edb50d0fe4720c5bdaf7535 (commit) from 2bc967ea2524a1a93cc596863cee1e156e619afe (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 30af1a8132ee1a74c890e9a4d7d2302e3715e14d Author: Phil Carns <carns(a)mcs.anl.gov> Date: Tue Mar 20 12:06:22 2012 -0400 starting point for idb-style tosd benchmark - incomplete right now commit b88b17823f9e28705edb50d0fe4720c5bdaf7535 Author: Phil Carns <carns(a)mcs.anl.gov> Date: Tue Mar 20 12:05:00 2012 -0400 add args to init fn on aesop_main_set_with_init ----------------------------------------------------------------------- Summary of changes: code/src/aesop/resource.h | 4 +- code/src/transactional-osd/tests/module.mk.in | 1 + code/src/transactional-osd/tests/tosd-bench-idb.ae | 243 ++++++++++++++++++++ code/src/transactional-osd/tests/tosd-niid.ae | 2 +- 4 files changed, 247 insertions(+), 3 deletions(-) create mode 100644 code/src/transactional-osd/tests/tosd-bench-idb.ae Diff of changes: diff --git a/code/src/aesop/resource.h b/code/src/aesop/resource.h index 2be890c..3ea4f55 100644 --- a/code/src/aesop/resource.h +++ b/code/src/aesop/resource.h @@ -227,10 +227,10 @@ int main(int argc, char **argv) \ ae_hints_t __main_hints; \ ae_op_id_t __main_opid; \ ae_ret_t ret; \ - ae_ret_t (*__main_init_func)(void) = __init_before_main__; \ + ae_ret_t (*__main_init_func)(int argc, char** argv) = __init_before_main__; \ if(__init_before_main__) \ { \ - ret = __main_init_func(); \ + ret = __main_init_func(argc, argv); \ aesop_error_assert(ret); \ } \ ret = aesop_module_init(__init_module__); \ diff --git a/code/src/transactional-osd/tests/module.mk.in b/code/src/transactional-osd/tests/module.mk.in index 812c477..84c92ab 100644 --- a/code/src/transactional-osd/tests/module.mk.in +++ b/code/src/transactional-osd/tests/module.mk.in @@ -9,6 +9,7 @@ AETESTSRC += $(DIR)/tosd-niid.ae AETESTSRC += $(DIR)/fake1.ae AETESTSRC += $(DIR)/tosd-multi-object.ae AETESTSRC += $(DIR)/tosd-bench-concurrent.ae +AETESTSRC += $(DIR)/tosd-bench-idb.ae AETESTSRC += $(DIR)/tosd-bench-create.ae AETESTSRC += $(DIR)/workload-gen.ae AESOP_HDR += $(DIR)/fake-resource.hae diff --git a/code/src/transactional-osd/tests/tosd-bench-idb.ae b/code/src/transactional-osd/tests/tosd-bench-idb.ae new file mode 100644 index 0000000..579cdcb --- /dev/null +++ b/code/src/transactional-osd/tests/tosd-bench-idb.ae @@ -0,0 +1,243 @@ + +#include <stdio.h> +#include <errno.h> +#include <sys/time.h> +#include <stdlib.h> +#include <limits.h> +#include "src/aesop/aesop.h" +#include "src/transactional-osd/transactional-osd.hae" +#include "src/zeroconf/zeroconf.h" + +enum op_type +{ + WRITE, + READ +}; + +struct bench_op +{ + int64_t offset; + int64_t size; + enum op_type type; + struct triton_list_link list_link; +}; + +TRITON_LIST_DEFINE(op_list); +triton_mutex_t op_list_mutex = TRITON_MUTEX_INITIALIZER; + +static double Wtime(void) +{ + struct timeval t; + gettimeofday(&t, NULL); + return((double)t.tv_sec + (double)(t.tv_usec) / 1000000); +} +static double start_tm; +static double end_tm; +static int concurrent; +static uint64_t first_version; +static uint64_t version; +static int max_seconds = 0; +static uint128_t oid; + +static __blocking int do_tosd_test(void) +{ + triton_ret_t tret; + int w_flags = TOSD_FLAG_AUTO_TXN; + + tret = tosd_create(oid, 0); + if(tret == TRITON_ERR_EXIST) + { + printf("# operating on existing object.\n"); + triton_error_destroy(tret); + tret = TRITON_SUCCESS; + } + triton_error_assert(tret); + + tret = tosd_get_version(oid, &version); + triton_error_assert(tret); + + first_version = version; + + start_tm = Wtime(); + + pwait + { + pprivate int i; + pprivate struct triton_list_link* tmp_link; + pprivate char* buffer = NULL; + pprivate uint64_t b_version; + pprivate struct bench_op* tmp_op; + pprivate triton_ret_t b_ret; + pprivate int64_t out_size; + pprivate int64_t keyval_offset = 0; + pprivate int64_t* offset_ptr; + pprivate uint64_t fork; + pprivate int ret; + pprivate double now; + + for(i=0; i<concurrent; i++) + { + pbranch + { + triton_mutex_lock(&op_list_mutex); + tmp_link = triton_stack_pop(&op_list); + triton_mutex_unlock(&op_list_mutex); + + while(tmp_link) + { + b_version = version++; + + tmp_op = triton_list_get_entry(tmp_link, struct bench_op, + list_link); + + offset_ptr = &tmp_op->offset; + fork = 0; + + if(tmp_op->type == WRITE) + { + b_ret = tosd_write(oid, fork, b_version, &buffer, + &tmp_op->size, 1, &tmp_op->offset, + &tmp_op->size, 1, w_flags, 0); + triton_error_assert(b_ret); + } + else + { + assert(tmp_op->type == READ); + + b_ret = tosd_read(oid, fork, &buffer, &tmp_op->size, 1, + &tmp_op->offset, &tmp_op->size, 1, &out_size, 0); + triton_error_assert(b_ret); + assert(out_size == tmp_op->size); + } + + now = Wtime(); + triton_mutex_lock(&op_list_mutex); + /* check the time limit if applicable */ + if(max_seconds > 0 && (now - start_tm) > max_seconds) + tmp_link = NULL; + else + tmp_link = triton_stack_pop(&op_list); + triton_mutex_unlock(&op_list_mutex); + } + } + } + } + + end_tm = Wtime(); + + return 0; +} + +__blocking int aesop_main(int argc, char **argv) +{ + ae_context_t ctx; + ae_hints_t hints; + ae_op_id_t op_id; + int pc = 0; + FILE *desc = 0; + char line[2048]; + char op_string[100]; + int64_t size; + int64_t offset; + int ret; + struct bench_op* tmp_op; + triton_ret_t tret; + const char* tosd_init_flags = NULL; + int i; + + oid = triton_uint128_from_uint64(1); + + if(argc != 5) + { + fprintf(stderr, "Usage: tosd-bench-concurrent <workload description file> <dir> <concurrency> <threaded y|n>\n"); + return(-1); + } + + ret = sscanf(argv[3], "%d", &concurrent); + if(ret != 1 || concurrent < 1) + { + fprintf(stderr, "Usage: tosd-bench-concurrent <workload description file> <dir> <concurrency> <threaded y|n>\n"); + return(-1); + } + + /* parse description of workload */ + desc = fopen(argv[1], "r"); + if(!desc) + { + perror("fopen"); + return(-1); + } + while(fgets(line, 2048, desc)) + { + if(line[0] == '#') + continue; +#if SIZEOF_LONG_INT == 4 + ret = sscanf(line, "%s %lld %lld", op_string, &offset, &size); +#else + ret = sscanf(line, "%s %ld %ld", op_string, &offset, &size); +#endif + if(ret != 3) + { + fprintf(stderr, "Error: bad line: %s\n", line); + return(-1); + } + tmp_op = malloc(sizeof(*tmp_op)); + assert(tmp_op); + + if(strcmp(op_string, "write") == 0) + tmp_op->type = WRITE; + else if(strcmp(op_string, "read") == 0) + tmp_op->type = READ; + else + assert(0); + tmp_op->offset = offset; + tmp_op->size = size; + triton_list_add_back(&tmp_op->list_link, &op_list); + } + fclose(desc); + + do_tosd_test(); + + /* print command line in comments */ + printf("#"); + for(i=0; i<argc; i++) + { + printf(" %s", argv[i]); + } + printf("\n"); + printf("# <seconds> <ops/s>\n"); + printf("%f\t%f\n", end_tm-start_tm, ((double)(version-first_version))/(end_tm-start_tm)); + + return 0; +} + +static triton_ret_t init_tosd_zeroconf_params(int argc, char** argv) +{ + triton_ret_t tret; + char tosd_path_logs[PATH_MAX]; + + tret = triton_zeroconf_set("triton.tosd.db_path", argv[2]); + triton_error_assert(tret); + + sprintf(tosd_path_logs, "%s/logs", argv[2]); + tret = triton_zeroconf_set("triton.tosd.data_path", tosd_path_logs); + triton_error_assert(tret); + + tret = triton_zeroconf_set("triton.tosd.alignment", "1"); + triton_error_assert(tret); + + return(TRITON_SUCCESS); +} + +aesop_main_set_with_init(init_tosd_zeroconf_params, + "triton.tosd", + aesop_main); + +/* + * Local variables: + * c-indent-level: 4 + * c-basic-offset: 4 + * End: + * + * vim: ft=c ts=8 sts=4 sw=4 expandtab + */ diff --git a/code/src/transactional-osd/tests/tosd-niid.ae b/code/src/transactional-osd/tests/tosd-niid.ae index 45d6736..b5db64d 100644 --- a/code/src/transactional-osd/tests/tosd-niid.ae +++ b/code/src/transactional-osd/tests/tosd-niid.ae @@ -125,7 +125,7 @@ static __blocking int do_tosd_test(int argc, char **argv) return 1; } -static triton_ret_t init_tosd_zeroconf_params(void) +static triton_ret_t init_tosd_zeroconf_params(int argc, char** argv) { triton_ret_t ret; hooks/post-receive -- Triton Repository

1 0

Triton-private Repository branch, master, updated. dea32ecc4756ced6307d7e9820858917f92aa043
by noreply＠mcs.anl.gov 16 Mar '12

16 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton-private Repository". The branch, master has been updated via dea32ecc4756ced6307d7e9820858917f92aa043 (commit) via 8d011bf3e088502fcf9e2976fba63098edecc39e (commit) via 0b7fdb8124b438fd179d53fb8973882a625d7524 (commit) from 643a7fae59bfbbf6d779a32161587f94b2233be6 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit dea32ecc4756ced6307d7e9820858917f92aa043 Author: Dries Kimpe <dkimpe(a)mcs.anl.gov> Date: Fri Mar 16 21:06:16 2012 -0500 Remove unused files commit 8d011bf3e088502fcf9e2976fba63098edecc39e Author: Dries Kimpe <dkimpe(a)mcs.anl.gov> Date: Fri Mar 16 21:05:40 2012 -0500 Some notes commit 0b7fdb8124b438fd179d53fb8973882a625d7524 Author: Dries Kimpe <dkimpe(a)mcs.anl.gov> Date: Fri Mar 16 20:57:39 2012 -0500 Minor fixes. Final version. ----------------------------------------------------------------------- Summary of changes: papers/2011/aesop/TO_IMPROVE | 20 ++++++ papers/2011/aesop/evaluation.tex | 22 ++++--- papers/2011/aesop/intro.tex | 4 +- papers/2011/aesop/listings/listing-1.ae | 22 ------- papers/2011/aesop/listings/listing-1.ae.tex | 32 ---------- papers/2011/aesop/script2tex.pl | 84 --------------------------- 6 files changed, 35 insertions(+), 149 deletions(-) create mode 100644 papers/2011/aesop/TO_IMPROVE delete mode 100644 papers/2011/aesop/listings/listing-1.ae delete mode 100644 papers/2011/aesop/listings/listing-1.ae.tex delete mode 100755 papers/2011/aesop/script2tex.pl Diff of changes: diff --git a/papers/2011/aesop/TO_IMPROVE b/papers/2011/aesop/TO_IMPROVE new file mode 100644 index 0000000..043fd0a --- /dev/null +++ b/papers/2011/aesop/TO_IMPROVE @@ -0,0 +1,20 @@ +Things we can improve: + +- Evaluation: + - A point we're trying to make is that we can retarget and still obtain optimal performance. + We don't show that. We don't show retargetting since we only run on one system. + Even more so, it seems that straight threads (per client or per op) perform very well; + + So, we map from linear flow to linear flow performance (as events perform badly on this system). + + - What's the point of publishing the -null graphs? We could drop them, and the paper would not become + weaker. Actually, to explain the performance difference, we even say that if you don't use + the blocking functions (write ipv async write), the performance drops to the aesop level. + + So, in other words, we claim we can retarget but even on our test system for the -null case, + we don't manage to reach optimal performance and as an explanation we're saying that it is because + the async API does not perform as well as the thread API? + + Wasn't that the point of aesop, to be able to retarget to the optimal API? + + diff --git a/papers/2011/aesop/evaluation.tex b/papers/2011/aesop/evaluation.tex index 67343d0..88b9932 100644 --- a/papers/2011/aesop/evaluation.tex +++ b/papers/2011/aesop/evaluation.tex @@ -125,12 +125,11 @@ on 65 nodes; one server node and 1-64 client nodes. \subsection{Run-time Efficiency} -Figure~\ref{fig:runtime} shows the results for the four runtime test types -with each plotted as the median of five samples with the error bars -represtenting the minimum and maximum rates. -Figure~\ref{fig:write} shows the request servicing rate of the concurrent write -workload for each server implementation as the number of client processes -is scaled from 16 to 1,024. In this graph we see that {\aesop} does not +Figure~\ref{fig:runtime} shows, for each of the four test types, +the median, minimum, and maximum number of requests completed per second. Each +test type was executed 5 times, with the number of clients ranging from 16 to +1,024. +In Figure~\ref{fig:write} we see that {\aesop} does not perform as well as the other servers for a small number of clients (delivering 79 operations per second at the smallest scale, versus 132 ops/s for the thread-per-op server). However, {\aesop} is the fastest server at the largest @@ -171,11 +170,16 @@ system. Another notable observation in these graphs is that the \aesop server is competitive at small scale and, in fact, is the fastest implementation in the 16-client-process read-null test and nearly the fastest in the -16-client-process write-null test. These results support the observation from the previous -section that poor \aesop performance at small scale is likely a tuning flaw -in the file resource used in the \aesop standard library, rather than a +16-client-process write-null test. These results support the observation + that poor \aesop performance at small scale is likely a tuning flaw +in the \aesop implementation of I/O functionality, rather than a fundamental programming language problem. +%However, Figure~\ref{fig:runtime} shows the results obtained on a single +%system. On a different system the event server, which showed the least The main goal of performing this evaluation is to determine if it is +%indeed possible to tune \aesop to, + + The test client also measures the latency of each individual request and then computes the minimum and maximum latency, the first quartile latency, and the third diff --git a/papers/2011/aesop/intro.tex b/papers/2011/aesop/intro.tex index 4d37cc8..ec4fa03 100644 --- a/papers/2011/aesop/intro.tex +++ b/papers/2011/aesop/intro.tex @@ -100,8 +100,8 @@ core algorithm description is decoupled from these architecture-specific details, the run-time system can be tuned to match different system architectures, different device APIs, and different workloads without any change to the core system software algorithms. Unlike event-driven -architectures, the \aesop extensions do not suffer from an inversion of -control, preserving readable control flow regardless of how many steps or +architectures, +\aesop preserves readable control flow regardless of how many steps or concurrent devices are utilized in a given algorithm. \aesop also offers functionality that is not available in a traditional multithreaded or event-driven framework -- for example the ability to cleanly cancel outstanding diff --git a/papers/2011/aesop/listings/listing-1.ae b/papers/2011/aesop/listings/listing-1.ae deleted file mode 100644 index 0a7977f..0000000 --- a/papers/2011/aesop/listings/listing-1.ae +++ /dev/null @@ -1,22 +0,0 @@ -aesop void doWrite (...) { - pwait { - pbranch { - aesop_timer (TIMEOUT); - ae_cancel_branches (); - } - pbranch { - replicateWrite (...); - ae_cancel_branches (); - } - } -} - -aesop void replicateWrite (...) { - pwait { - pprivate int i; - for (i=0; i<REPLICAS; ++i) - pbranch { - replicateWriteTo (i); - } - } -} diff --git a/papers/2011/aesop/listings/listing-1.ae.tex b/papers/2011/aesop/listings/listing-1.ae.tex deleted file mode 100644 index 238a1b7..0000000 --- a/papers/2011/aesop/listings/listing-1.ae.tex +++ /dev/null @@ -1,32 +0,0 @@ - -\begin{center} -\scriptsize - -\fbox{ -\begin{tabular}{r|l} - 1 & {\tt aesop\ void\ doWrite\ (...)\ \{ } \\ - 2 & {\tt \ \ pwait\ \{ } \\ - 3 & {\tt \ \ \ \ pbranch\ \{ } \\ - 4 & {\tt \ \ \ \ \ \ aesop\_timer\ (TIMEOUT); } \\ - 5 & {\tt \ \ \ \ \ \ ae\_cancel\_branches\ (); } \\ - 6 & {\tt \ \ \ \ \} } \\ - 7 & {\tt \ \ \ \ pbranch\ \{ } \\ - 8 & {\tt \ \ \ \ \ \ replicateWrite\ (...); } \\ - 9 & {\tt \ \ \ \ \ \ ae\_cancel\_branches\ (); } \\ -10 & {\tt \ \ \ \ \} } \\ -11 & {\tt \ \ \} } \\ -12 & {\tt \} } \\ -13 & {\tt } \\ -14 & {\tt aesop\ void\ replicateWrite\ (...)\ \{ } \\ -15 & {\tt \ \ pwait\ \{ } \\ -16 & {\tt \ \ \ \ pprivate\ int\ i; } \\ -17 & {\tt \ \ \ \ for\ (i=0;\ i<REPLICAS;\ ++i) } \\ -18 & {\tt \ \ \ \ \ \ pbranch\ \{ } \\ -19 & {\tt \ \ \ \ \ \ \ \ replicateWriteTo\ (i); } \\ -20 & {\tt \ \ \ \ \ \ \} } \\ -21 & {\tt \ \ \} } \\ -22 & {\tt \} } \\ - -\end{tabular} -} -\end{center} diff --git a/papers/2011/aesop/script2tex.pl b/papers/2011/aesop/script2tex.pl deleted file mode 100755 index 20ef069..0000000 --- a/papers/2011/aesop/script2tex.pl +++ /dev/null @@ -1,84 +0,0 @@ -#!/usr/bin/perl -n - -# Process code for insertion into LaTeX so verbatim is not required - -# example usage: script2tex.pl file.c > file.c.tex - -# Makes a nice line-numbered table -# Uses \tt -# Escapes curly braces -# Escapes percent sign (%) -# Allows for bold sections -# Use _B_ to get {\bf (BOLD) -# Use _C_ to get } (CLOSE BRACE) -# Escapes underscores -# Escapes spaces (for indentation) -# Puts C++ comments in textrm -# Puts shell (hash) comments in textrm -# Use "-Example- 32" to get a label - -# If NO_BOX is non-empty in the environment, suppress the fbox - -BEGIN { - $N = 1; - $EXAMPLE = -1; - $BOX = 1; - if ($ENV{NO_BOX} != "") { - $BOX = 0; - } - - print "\n"; - print "\\begin{center}\n"; - print "\\scriptsize\n\n"; - if ($BOX == 1) { - print "\\fbox{\n"; - } - print "\\begin{tabular}{r|l}\n"; -} - -{ - chomp; - if (/.*-Example- (.*)/) { - $EXAMPLE = $1; - } - else { - printf("%2i & ", $N); - printf("{\\tt "); - s/{/\\{/g; - s/}/\\}/g; - s/%/\\%/g; - s{//(.*)}{// \\textrm{$1}}g; - s{#(.*)}{\\# \\textrm{$1}}g; - s/ /\\ /g; - s/_B_/{\\bf /g; - s/_C_/}/g; - s/_/\\_/g; - printf("%-40s", $_); - printf("} "); - printf("\\\\\n"); - $N++; - } -} - -END { - print "\n"; - if ($EXAMPLE != -1) { - print "\\hline\n"; - } - print "\\end{tabular}\n"; - if ($BOX == 1) { - print "}\n"; - } - if ($EXAMPLE != -1) { - print "\\\\\n"; - print "\\vspace{1mm}\n"; - print "\\textbf{Example $EXAMPLE}\n"; - } - print "\\end{center}\n"; -} - -# Local Variables: -# indent-tabs-mode: f -# tab-width: 2 -# perl-indent-level: 2 -# End: hooks/post-receive -- Triton-private Repository

1 0

Triton-private Repository branch, master, updated. 643a7fae59bfbbf6d779a32161587f94b2233be6
by noreply＠mcs.anl.gov 16 Mar '12

16 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton-private Repository". The branch, master has been updated via 643a7fae59bfbbf6d779a32161587f94b2233be6 (commit) from 3fb8275927df261edb91ba038be84b45965ada61 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 643a7fae59bfbbf6d779a32161587f94b2233be6 Author: Kevin Harms <harms(a)alcf.anl.gov> Date: Fri Mar 16 17:59:00 2012 -0500 Remove the paragraph in the related work comparing aesop and openmp and merge it with the above paragraph that was doing the same thing. ----------------------------------------------------------------------- Summary of changes: papers/2011/aesop/related.tex | 34 +++++++++++++++++++--------------- 1 files changed, 19 insertions(+), 15 deletions(-) Diff of changes: diff --git a/papers/2011/aesop/related.tex b/papers/2011/aesop/related.tex index b7b4474..148d18d 100644 --- a/papers/2011/aesop/related.tex +++ b/papers/2011/aesop/related.tex @@ -16,23 +16,27 @@ application level, rather than coordinating asynchronous activity on multiple I/O devices at the system software level. As a result, these languages optimize for maximum use of available CPU resources, whereas \aesop optimizes for maximum use of underlying devices that may or may not require -significant CPU activity. The level of \aesop concurrency is not -correlated with the number of CPU cores available, though \aesop can utilize -multiple cores (if available) to improve latency and concurrency. +significant CPU activity. \aesop in general has the closest similarity to +OpenMP but differs in that \aesop does not create or require the use of threads. +The level of \aesop concurrency is not correlated with the number of CPU +cores available, though \aesop can utilize multiple cores (if available) to +improve latency and concurrency. The other main differentiation is that \aesop +also offers language support for cleanly cancelling outstanding concurrent +executions paths. % this is repeated from above -\aesop has clear similarities to OpenMP, the C/Fortran language extensions for -expressing thread parallelism. In particular \aesop supports decoupling -sequential control flow (i.e., parallel for), waiting for their completion -(barrier) and the annotation of variables to indicate whether they are to be shared -or private with respect to concurrent execution flows. Since all these concepts -are present in OpenMP as well, it is easy to see how \aesop might be used to -write multithreaded applications. However, this is not the primary focus of -\aesop. \aesop does not create or require threads. While OpenMP mainly -provides a portable thread abstraction, it does not offer any support for -asynchronous device interfaces. \aesop also differs from the related work in -that only \aesop offers language support for cleanly cancelling outstanding -concurrent execution paths. +%\aesop has clear similarities to OpenMP, the C/Fortran language extensions for +%expressing thread parallelism. In particular \aesop supports decoupling +%sequential control flow (i.e., parallel for), waiting for their completion +%(barrier) and the annotation of variables to indicate whether they are to be shared +%or private with respect to concurrent execution flows. Since all these concepts +%are present in OpenMP as well, it is easy to see how \aesop might be used to +%write multithreaded applications. However, this is not the primary focus of +%\aesop. \aesop does not create or require threads. While OpenMP mainly +%provides a portable thread abstraction, it does not offer any support for +%asynchronous device interfaces. \aesop also differs from the related work in +%that only \aesop offers language support for cleanly cancelling outstanding +%concurrent execution paths. Grand Central Dispatch (GCD)~\cite{gcd-web} is a popular programming model introduced by Apple Inc. that allows developers to describe tasks to be hooks/post-receive -- Triton-private Repository

1 0

Triton-private Repository branch, master, updated. 3fb8275927df261edb91ba038be84b45965ada61
by noreply＠mcs.anl.gov 16 Mar '12

16 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton-private Repository". The branch, master has been updated via 3fb8275927df261edb91ba038be84b45965ada61 (commit) from ae57642029dd11b77dc55b8851d730fc6032eea1 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 3fb8275927df261edb91ba038be84b45965ada61 Author: Kevin Harms <harms(a)alcf.anl.gov> Date: Fri Mar 16 17:44:29 2012 -0500 Fix oops in comment block, add statement about errorbars. ----------------------------------------------------------------------- Summary of changes: papers/2011/aesop/evaluation.tex | 15 ++++++++------- 1 files changed, 8 insertions(+), 7 deletions(-) Diff of changes: diff --git a/papers/2011/aesop/evaluation.tex b/papers/2011/aesop/evaluation.tex index c8ccda3..67343d0 100644 --- a/papers/2011/aesop/evaluation.tex +++ b/papers/2011/aesop/evaluation.tex @@ -47,6 +47,7 @@ follows. \caption{Run-time performance for each test case.\\ \footnotesize (X-axis is the number of client instances, each graph has a distinct Y-axis.)} + \label{fig:runtime} \vspace{-.2in} \end{figure*} @@ -108,7 +109,7 @@ to be the aggregate run time. Each client also tracks the time needed to servic each individual request in order to analyze individual request latency. The client test harness was configured to generate requests of size 4 KiB in all cases. Each client process performed 16 -requests in the write and read experiments and 4,096 requests in the +requests in the write and read tests and 4,096 requests in the write-null and read-null tests. All experiments were executed on the Fusion cluster in the Argonne @@ -124,6 +125,9 @@ on 65 nodes; one server node and 1-64 client nodes. \subsection{Run-time Efficiency} +Figure~\ref{fig:runtime} shows the results for the four runtime test types +with each plotted as the median of five samples with the error bars +represtenting the minimum and maximum rates. Figure~\ref{fig:write} shows the request servicing rate of the concurrent write workload for each server implementation as the number of client processes is scaled from 16 to 1,024. In this graph we see that {\aesop} does not @@ -254,14 +258,11 @@ event & 28 & 23 & 341 \\ %\subsubsection{Quantitative} -We approximate programmer productivity by measuring code complexity. +We infer programmer productivity by measuring code complexity. Table~\ref{tab:complexity} compares the code complexity of each server implemenation using McCabe Cyclomatic Complexity (CC)~\cite{mccabe}, -Modified McCabe Cyclomatic Complexity (mod. CC), and -Source Lines of Code (SLOC). The CC and Mod. CC metrics were measured by -using the pmccabe tool, version 2.6, created by Paul Bame~\cite{pmccabe}; -the SLOC metrics were measured using the sloccount tool, -version 2.26, created by David A. Wheeler\begin{comment}~\cite{sloccount}\end{comment}. +Modified McCabe Cyclomatic Complexity (mod. CC) using \emph{pmccabe} v2.6 by +Paul Bame, and Source Lines of Code (SLOC) using \emph{sloccount} v2.26 by David A. Wheeler. In order to simplify the comparison, all four servers had no error handling except for assertions on expected return codes. The protocol definition as well as helper functions for sending hooks/post-receive -- Triton-private Repository

1 0

Triton-private Repository branch, master, updated. ae57642029dd11b77dc55b8851d730fc6032eea1
by noreply＠mcs.anl.gov 16 Mar '12

16 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton-private Repository". The branch, master has been updated via ae57642029dd11b77dc55b8851d730fc6032eea1 (commit) from 4b37dcaef317530f93fd30975108013eee018db7 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit ae57642029dd11b77dc55b8851d730fc6032eea1 Author: Dries Kimpe <dkimpe(a)mcs.anl.gov> Date: Fri Mar 16 17:26:51 2012 -0500 fix ----------------------------------------------------------------------- Summary of changes: papers/2011/aesop/intro.tex | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) Diff of changes: diff --git a/papers/2011/aesop/intro.tex b/papers/2011/aesop/intro.tex index 1bf6766..4d37cc8 100644 --- a/papers/2011/aesop/intro.tex +++ b/papers/2011/aesop/intro.tex @@ -89,7 +89,7 @@ engineering challenges associated with developing distributed system software. Its goal is to improve programmer productivity by presenting a convenient model for developing distributed system software. The model isolates the developer from how concurrency is managed, but maintains execution efficiency -by being able to map \aesop code to the optimal concurrency model required by the +through its ability to map \aesop code to the concurrency model preferred by the system or device. \aesop is a superset of the C language. It describes blocks of code that can hooks/post-receive -- Triton-private Repository

1 0

Triton-private Repository branch, master, updated. 4b37dcaef317530f93fd30975108013eee018db7
by noreply＠mcs.anl.gov 16 Mar '12

16 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton-private Repository". The branch, master has been updated via 4b37dcaef317530f93fd30975108013eee018db7 (commit) from 0dc90853411b178994027c28142d26b475a9d018 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 4b37dcaef317530f93fd30975108013eee018db7 Author: Justin M Wozniak <wozniak(a)mcs.anl.gov> Date: Fri Mar 16 17:04:06 2012 -0500 Add initial to active author list ----------------------------------------------------------------------- Summary of changes: papers/2011/aesop/aesop.tex | 2 +- papers/2011/aesop/model.tex | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) Diff of changes: diff --git a/papers/2011/aesop/aesop.tex b/papers/2011/aesop/aesop.tex index f069603..7620496 100644 --- a/papers/2011/aesop/aesop.tex +++ b/papers/2011/aesop/aesop.tex @@ -97,7 +97,7 @@ xleftmargin=6ex % use a multiple column layout for up to two different % affiliations -\author{\IEEEauthorblockN{Dries Kimpe, Philip Carns, Kevin Harms, Justin +\author{\IEEEauthorblockN{Dries Kimpe, Philip Carns, Kevin Harms, Justin M. Wozniak, Samuel Lang, Robert Ross} \IEEEauthorblockA{Mathematics and Computer Science Division\\ Argonne National Laboratory, Argonne, IL 60439\\ diff --git a/papers/2011/aesop/model.tex b/papers/2011/aesop/model.tex index 5d54492..31d4a73 100644 --- a/papers/2011/aesop/model.tex +++ b/papers/2011/aesop/model.tex @@ -20,7 +20,7 @@ parallel for), in that no statements following the pwait will execute unless all the pbranches it encloses have completed. \aesop also supports {\em lonely pbranches}, that is, pbranches created outside a pwait. A lonely pbranch resembles a POSIX detached thread in that -lonely pbranches cannot be synchronized with. +lonely pbranches cannot be synchronized with other user code. By default, variables in \aesop are shared between pbranches. This configuration is similar hooks/post-receive -- Triton-private Repository

1 0

Triton-private Repository branch, master, updated. 0dc90853411b178994027c28142d26b475a9d018
by noreply＠mcs.anl.gov 16 Mar '12

16 Mar '12

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Triton-private Repository". The branch, master has been updated via 0dc90853411b178994027c28142d26b475a9d018 (commit) from 21370ccd7f4674b848e98e7ff895ca7bfe1b6dff (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 0dc90853411b178994027c28142d26b475a9d018 Author: Justin M Wozniak <wozniak(a)mcs.anl.gov> Date: Fri Mar 16 17:00:23 2012 -0500 BDB fix ----------------------------------------------------------------------- Summary of changes: papers/2011/aesop/aesop.bib | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) Diff of changes: diff --git a/papers/2011/aesop/aesop.bib b/papers/2011/aesop/aesop.bib index 5962897..f2be741 100644 --- a/papers/2011/aesop/aesop.bib +++ b/papers/2011/aesop/aesop.bib @@ -361,7 +361,7 @@ address = {Los Alamitos, CA, USA}, @INPROCEEDINGS{BDB_1999, author = {Olson, Michael A. and Bostic, Keith and Seltzer, Margo}, - title = {Berkeley DB}, + title = {Berkeley {DB}}, booktitle = {Proceedings of USENIX}, year = {1999}, } hooks/post-receive -- Triton-private Repository

1 0