summaryrefslogtreecommitdiffstats
path: root/benchmarks
diff options
context:
space:
mode:
authorDouglas Rumbaugh <dbr4@psu.edu>2025-06-02 19:54:15 -0400
committerDouglas Rumbaugh <dbr4@psu.edu>2025-06-02 19:54:15 -0400
commit6088ede2b5a8f580f78d4422162d423a7a11f520 (patch)
tree9124cab872af7a02209c3af758d4a4bf95b1aa26 /benchmarks
parent7f0acb5dc66cc93045ab0e40ebd5570388f9102b (diff)
downloaddynamic-extension-6088ede2b5a8f580f78d4422162d423a7a11f520.tar.gz
updates
Diffstat (limited to 'benchmarks')
-rw-r--r--benchmarks/tail-latency/btree_insert_dist.cpp2
-rw-r--r--benchmarks/tail-latency/fixed_shard_count.cpp19
-rw-r--r--benchmarks/tail-latency/insert_query_threads.cpp4
-rw-r--r--benchmarks/tail-latency/isam_insert_dist.cpp107
-rw-r--r--benchmarks/tail-latency/knn_query_parm_sweep.cpp14
-rw-r--r--benchmarks/tail-latency/query_parm_sweep.cpp8
-rw-r--r--benchmarks/tail-latency/selectivity_sweep.cpp4
-rw-r--r--benchmarks/tail-latency/standard_latency_dist.cpp18
8 files changed, 39 insertions, 137 deletions
diff --git a/benchmarks/tail-latency/btree_insert_dist.cpp b/benchmarks/tail-latency/btree_insert_dist.cpp
index af60819..e1d171d 100644
--- a/benchmarks/tail-latency/btree_insert_dist.cpp
+++ b/benchmarks/tail-latency/btree_insert_dist.cpp
@@ -45,7 +45,7 @@ int main(int argc, char **argv) {
auto queries = read_range_queries<QP>(q_fname, .0001);
/* warmup structure w/ 10% of records */
- size_t warmup = .1 * n;
+ size_t warmup = .3 * n;
for (size_t i=0; i<warmup; i++) {
btree.insert(data[i]);
}
diff --git a/benchmarks/tail-latency/fixed_shard_count.cpp b/benchmarks/tail-latency/fixed_shard_count.cpp
index e980bcf..45c62fb 100644
--- a/benchmarks/tail-latency/fixed_shard_count.cpp
+++ b/benchmarks/tail-latency/fixed_shard_count.cpp
@@ -2,6 +2,7 @@
*
*/
+#include "framework/scheduling/SerialScheduler.h"
#define ENABLE_TIMER
#define TS_TEST
@@ -9,8 +10,9 @@
#include "framework/DynamicExtension.h"
#include "framework/scheduling/FIFOScheduler.h"
-#include "shard/TrieSpline.h"
+#include "shard/ISAMTree.h"
#include "query/rangecount.h"
+#include "framework/util/Configuration.h"
#include "framework/interface/Record.h"
#include "file_util.h"
#include "standard_benchmarks.h"
@@ -23,10 +25,11 @@
typedef de::Record<uint64_t, uint64_t> Rec;
-typedef de::TrieSpline<Rec> Shard;
+typedef de::ISAMTree<Rec> Shard;
typedef de::rc::Query<Shard> Q;
-typedef de::DynamicExtension<Shard, Q, de::DeletePolicy::TOMBSTONE, de::FIFOScheduler> Ext;
+typedef de::DynamicExtension<Shard, Q, de::DeletePolicy::TOMBSTONE, de::SerialScheduler> Ext;
typedef Q::Parameters QP;
+typedef de::DEConfiguration<Shard, Q, de::DeletePolicy::TOMBSTONE, de::SerialScheduler> Conf;
void usage(char *progname) {
fprintf(stderr, "%s reccnt datafile queryfile\n", progname);
@@ -52,10 +55,16 @@ int main(int argc, char **argv) {
for (size_t i=0; i<shard_counts.size(); i++) {
auto policy = get_policy<Shard, Q>(shard_counts[i], buffer_size, 4, n);
- auto extension = new Ext(std::move(policy));
+ auto config = Conf(std::move(policy));
+ config.recon_enable_maint_on_flush = false;
+ config.recon_maint_disabled = true;
+ config.buffer_flush_trigger = 4000;
+ config.maximum_threads = 8;
+
+ auto extension = new Ext(std::move(config));
/* warmup structure w/ 10% of records */
- size_t warmup = .1 * n;
+ size_t warmup = .3 * n;
for (size_t i=0; i<warmup; i++) {
while (!extension->insert(data[i])) {
usleep(1);
diff --git a/benchmarks/tail-latency/insert_query_threads.cpp b/benchmarks/tail-latency/insert_query_threads.cpp
index 3f78f16..cb4b81b 100644
--- a/benchmarks/tail-latency/insert_query_threads.cpp
+++ b/benchmarks/tail-latency/insert_query_threads.cpp
@@ -106,11 +106,11 @@ int main(int argc, char **argv) {
auto queries = read_sosd_point_lookups<QP>(q_fname, 100);
size_t buffer_size = 8000;
- std::vector<size_t> policies = {6};
+ std::vector<size_t> policies = {0};
std::vector<size_t> thread_counts = {32};
std::vector<size_t> modifiers = {0};
std::vector<size_t> scale_factors = {6};
- std::vector<double> rate_limits = {1, 0.9999, 0.999, 0.99, 0.9, 0.85, 0.8};
+ std::vector<double> rate_limits = {1};
size_t insert_threads = 1;
size_t query_threads = 1;
diff --git a/benchmarks/tail-latency/isam_insert_dist.cpp b/benchmarks/tail-latency/isam_insert_dist.cpp
deleted file mode 100644
index 88d37c5..0000000
--- a/benchmarks/tail-latency/isam_insert_dist.cpp
+++ /dev/null
@@ -1,107 +0,0 @@
-/*
- *
- */
-
-#include "framework/scheduling/FIFOScheduler.h"
-#define ENABLE_TIMER
-#define TS_TEST
-
-#include <thread>
-
-#include "framework/DynamicExtension.h"
-#include "shard/ISAMTree.h"
-#include "query/rangecount.h"
-#include "framework/interface/Record.h"
-#include "file_util.h"
-#include "standard_benchmarks.h"
-
-#include <gsl/gsl_rng.h>
-
-#include "psu-util/timer.h"
-
-
-typedef de::Record<uint64_t, uint64_t> Rec;
-typedef de::ISAMTree<Rec>Shard;
-typedef de::rc::Query<Shard> Q;
-typedef de::DynamicExtension<Shard, Q, de::DeletePolicy::TOMBSTONE, de::FIFOScheduler> Ext;
-typedef Q::Parameters QP;
-
-void usage(char *progname) {
- fprintf(stderr, "%s reccnt datafile queryfile\n", progname);
-}
-
-int main(int argc, char **argv) {
-
- if (argc < 4) {
- usage(argv[0]);
- exit(EXIT_FAILURE);
- }
-
- size_t n = atol(argv[1]);
- std::string d_fname = std::string(argv[2]);
- std::string q_fname = std::string(argv[3]);
-
-
- auto data = read_sosd_file<Rec>(d_fname, n);
- auto queries = read_range_queries<QP>(q_fname, .0001);
-
- std::vector<int> policies = {3};
- std::vector<size_t> buffers = {8000, 16000, 32000};
- std::vector<size_t> sfs = {8};
-
- for (size_t l=0; l<policies.size(); l++) {
- for (size_t j=0; j<buffers.size(); j++) {
- for (size_t k=0; k<sfs.size(); k++) {
- auto policy = get_policy<Shard, Q>(sfs[k], buffers[j], policies[l]);
- auto extension = new Ext(policy, buffers[j]/4, buffers[j]);
-
- /* warmup structure w/ 10% of records */
- size_t warmup = .1 * n;
- for (size_t i=0; i<warmup; i++) {
- while (!extension->insert(data[i])) {
- usleep(1);
- }
- }
-
- extension->await_next_epoch();
-
- TIMER_INIT();
-
- for (size_t i=warmup; i<data.size(); i++) {
- TIMER_START();
- while (!extension->insert(data[i])) {
- usleep(1);
- }
- TIMER_STOP();
-
- //fprintf(stdout, "%ld\t%ld\t%d\t%ld\n", sfs[k], buffers[j], policies[l], TIMER_RESULT());
- }
-
- extension->await_next_epoch();
-
- /* repeat the queries a bunch of times */
- for (size_t l=0; l<10; l++) {
- for (size_t i=0; i<queries.size(); i++) {
- TIMER_START();
- auto q = queries[i];
- auto res = extension->query(std::move(q));
- res.get();
- TIMER_STOP();
-
- fprintf(stdout, "%ld\t%ld\t%d\t%ld\n", sfs[k], buffers[j], policies[l], TIMER_RESULT());
- }
- }
-
-
- QP p = {0, 10000};
- auto res =extension->query(std::move(p));
-
- fprintf(stderr, "%ld\n", res.get());
- extension->await_next_epoch();
- delete extension;
- }}}
-
-
- fflush(stderr);
-}
-
diff --git a/benchmarks/tail-latency/knn_query_parm_sweep.cpp b/benchmarks/tail-latency/knn_query_parm_sweep.cpp
index dfd713d..1766c74 100644
--- a/benchmarks/tail-latency/knn_query_parm_sweep.cpp
+++ b/benchmarks/tail-latency/knn_query_parm_sweep.cpp
@@ -105,14 +105,14 @@ int main(int argc, char **argv) {
//auto queries =read_sosd_point_lookups<QP>(q_fname, 1);
auto data = read_vector_file<Rec, W2V_SIZE>(d_fname, n);
- auto queries = read_knn_queries<QP>(q_fname, 15, 1);
+ auto queries = read_knn_queries<QP>(q_fname, 100, 1);
- size_t buffer_size = 8000;
- std::vector<size_t> policies = {0};
+ size_t buffer_size = 1000;
+ std::vector<size_t> policies = {0, 1};
std::vector<size_t> thread_counts = {8};
std::vector<double> modifiers = {0};
- std::vector<size_t> scale_factors = {2, 4, 6, 8, 10};
+ std::vector<size_t> scale_factors = {2, 4, 6, 8, 16, 32, 128};
size_t insert_threads = 1;
size_t query_threads = 1;
@@ -171,14 +171,14 @@ int main(int argc, char **argv) {
extension->await_version();
/* run some queries to "warm up" the cache */
- for (size_t i=0; i<queries.size()*2; i++) {
+ for (size_t i=0; i<queries.size()*5; i++) {
auto q_idx = i % queries.size();
auto q = queries[q_idx];
auto res = extension->query(std::move(q)).get();
total_res.fetch_add(res.size());
}
- total_query_count.store(100000);
+ total_query_count.store(5000);
TIMER_INIT();
TIMER_START();
for (size_t i=0; i<total_query_count; i++) {
@@ -200,7 +200,7 @@ int main(int argc, char **argv) {
fprintf(stdout, "%ld\t%ld\t%ld\t%lf\t%ld\t%ld\t%ld\t%ld\n", internal_thread_cnt, pol, sf,
mod, extension->get_height(), extension->get_shard_count(),
insert_tput, query_lat);
- extension->print_scheduler_statistics();
+ //extension->print_scheduler_statistics();
//extension->print_scheduler_query_data();
//extension->print_structure();
fflush(stdout);
diff --git a/benchmarks/tail-latency/query_parm_sweep.cpp b/benchmarks/tail-latency/query_parm_sweep.cpp
index 36665a6..f2453aa 100644
--- a/benchmarks/tail-latency/query_parm_sweep.cpp
+++ b/benchmarks/tail-latency/query_parm_sweep.cpp
@@ -103,11 +103,11 @@ int main(int argc, char **argv) {
auto queries =read_sosd_point_lookups<QP>(q_fname, 1);
size_t buffer_size = 8000;
- std::vector<size_t> policies = {1};
+ std::vector<size_t> policies = {0, 1};
std::vector<size_t> thread_counts = {8};
std::vector<double> modifiers = {0};
- std::vector<size_t> scale_factors = {4, 4, 4, 4};
+ std::vector<size_t> scale_factors = {2, 4, 6, 8, 16, 32, 64};
size_t insert_threads = 1;
size_t query_threads = 1;
@@ -195,8 +195,8 @@ int main(int argc, char **argv) {
fprintf(stdout, "%ld\t%ld\t%ld\t%lf\t%ld\t%ld\t%ld\t%ld\n", internal_thread_cnt, pol, sf,
mod, extension->get_height(), extension->get_shard_count(),
insert_tput, query_lat);
- extension->print_scheduler_statistics();
- extension->print_scheduler_query_data();
+ //extension->print_scheduler_statistics();
+ //extension->print_scheduler_query_data();
//extension->print_structure();
fflush(stdout);
diff --git a/benchmarks/tail-latency/selectivity_sweep.cpp b/benchmarks/tail-latency/selectivity_sweep.cpp
index 77c088b..0fc0f42 100644
--- a/benchmarks/tail-latency/selectivity_sweep.cpp
+++ b/benchmarks/tail-latency/selectivity_sweep.cpp
@@ -49,13 +49,13 @@ int main(int argc, char **argv) {
auto data = read_sosd_file<Rec>(d_fname, n);
std::vector<std::vector<QP>> query_sets;
- std::vector<double> selectivities = {0, 0.00000001, 0.0000001, 0.000001, 0.00001, 0.0001, 0.001, 0.01, 0.1};
+ std::vector<double> selectivities = {0.0000001}; //, 0.000001, 0.00001, 0.0001, 0.001, 0.01, 0.1, .25};
for (auto sel: selectivities) {
query_sets.push_back(generate_uniform_range_queries<QP>(100, n, sel));
}
- std::vector<size_t> sfs = {2, 4, 8}; //, 4, 8, 16, 32, 64, 128, 256, 512, 1024};
+ std::vector<size_t> sfs = {2, 4, 8, 16, 32, 64, 128};
size_t buffer_size = 8000;
std::vector<size_t> policies = {0, 1};
diff --git a/benchmarks/tail-latency/standard_latency_dist.cpp b/benchmarks/tail-latency/standard_latency_dist.cpp
index 8f1594b..dca02bf 100644
--- a/benchmarks/tail-latency/standard_latency_dist.cpp
+++ b/benchmarks/tail-latency/standard_latency_dist.cpp
@@ -2,17 +2,18 @@
*
*/
+#define ENABLE_TIMER
+#define TS_TEST
+
#include "framework/scheduling/SerialScheduler.h"
#include "framework/util/Configuration.h"
#include "util/types.h"
-#define ENABLE_TIMER
-#define TS_TEST
#include <thread>
#include "framework/DynamicExtension.h"
#include "framework/scheduling/FIFOScheduler.h"
-#include "shard/TrieSpline.h"
+#include "shard/ISAMTree.h"
#include "query/rangecount.h"
#include "framework/interface/Record.h"
#include "file_util.h"
@@ -26,11 +27,11 @@
typedef de::Record<uint64_t, uint64_t> Rec;
-typedef de::TrieSpline<Rec> Shard;
+typedef de::ISAMTree<Rec> Shard;
typedef de::rc::Query<Shard> Q;
-typedef de::DynamicExtension<Shard, Q, de::DeletePolicy::TOMBSTONE, de::FIFOScheduler> Ext;
+typedef de::DynamicExtension<Shard, Q, de::DeletePolicy::TOMBSTONE, de::SerialScheduler> Ext;
typedef Q::Parameters QP;
-typedef de::DEConfiguration<Shard, Q, de::DeletePolicy::TOMBSTONE, de::FIFOScheduler> Conf;
+typedef de::DEConfiguration<Shard, Q, de::DeletePolicy::TOMBSTONE, de::SerialScheduler> Conf;
void usage(char *progname) {
fprintf(stderr, "%s reccnt datafile queryfile\n", progname);
@@ -50,9 +51,9 @@ int main(int argc, char **argv) {
auto data = read_sosd_file<Rec>(d_fname, n);
auto queries = read_range_queries<QP>(q_fname, .0001);
- std::vector<size_t> sfs = {2, 3, 4, 5, 6, 7, 8}; //, 4, 8, 16, 32, 64, 128, 256, 512, 1024};
+ std::vector<size_t> sfs = {4, 8}; //, 4, 8, 16, 32, 64, 128, 256, 512, 1024};
size_t buffer_size = 8000;
- std::vector<size_t> policies = {0, 1, 2};
+ std::vector<size_t> policies = {0, 1};
for (auto pol: policies) {
for (size_t i=0; i<sfs.size(); i++) {
@@ -83,7 +84,6 @@ int main(int argc, char **argv) {
for (size_t j=warmup; j<data.size(); j++) {
while (!extension->insert(data[j])) {
usleep(1);
- fprintf(stderr, "insert blocked %ld\r", j);
}
}
TIMER_STOP();