dolysis
/
rocksdb
mirror of https://github.com/facebook/rocksdb.git
3
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Improve RangeDelAggregator benchmarks (#4395) 

Summary:
Improve time measurements for AddTombstones to only include the
call and not the VectorIterator setup. Also add a new
add_tombstones_per_run flag to call AddTombstones multiple times per
aggregator, which will help simulate more realistic workloads.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4395

Differential Revision: D9996811

Pulled By: abhimadan

fbshipit-source-id: 5865a95c323fbd9b3606493013664b4890fe5a02
This commit is contained in:
Abhishek Madan 2018-09-21 16:04:12 -07:00 committed by Facebook Github Bot
parent 04d373b260
commit 3c350a7cf0
1 changed files with 36 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

61
db/range_del_aggregator_bench.cc
View File

@ -52,6 +52,9 @@ DEFINE_int32(seed, 0, "random number generator seed");
DEFINE_int32(should_deletes_per_run, 1, "number of ShouldDelete calls per run");
DEFINE_int32(add_tombstones_per_run, 1,
"number of AddTombstones calls per run");
namespace {
struct Stats {
@ -65,8 +68,9 @@ std::ostream& operator<<(std::ostream& os, const Stats& s) {
fmt_holder.copyfmt(os);
os << std::left;
os << std::setw(25)
<< "AddTombstones: " << s.time_add_tombstones / (FLAGS_num_runs * 1.0e3)
os << std::setw(25) << "AddTombstones: "
<< s.time_add_tombstones /
(FLAGS_add_tombstones_per_run * FLAGS_num_runs * 1.0e3)
<< " us\n";
os << std::setw(25) << "ShouldDelete (first): "
<< s.time_first_should_delete / (FLAGS_num_runs * 1.0e3) << " us\n";
@ -133,17 +137,16 @@ struct TombstoneStartKeyComparator {
const Comparator* cmp;
};
void AddTombstones(RangeDelAggregator* range_del_agg,
const std::vector<PersistentRangeTombstone>& range_dels) {
std::unique_ptr<InternalIterator> MakeRangeDelIterator(
const std::vector<PersistentRangeTombstone>& range_dels) {
std::vector<std::string> keys, values;
for (const auto& range_del : range_dels) {
auto key_and_value = range_del.tombstone.Serialize();
keys.push_back(key_and_value.first.Encode().ToString());
values.push_back(key_and_value.second.ToString());
}
std::unique_ptr<test::VectorIterator> range_del_iter(
return std::unique_ptr<test::VectorIterator>(
new test::VectorIterator(keys, values));
range_del_agg->AddTombstones(std::move(range_del_iter));
}
// convert long to a big-endian slice key
@ -171,32 +174,40 @@ int main(int argc, char** argv) {
std::default_random_engine random_gen(FLAGS_seed);
std::normal_distribution<double> normal_dist(FLAGS_tombstone_width_mean,
FLAGS_tombstone_width_stddev);
std::vector<rocksdb::PersistentRangeTombstone> persistent_range_tombstones(
FLAGS_num_range_tombstones);
auto mode =
FLAGS_use_collapsed
? rocksdb::RangeDelPositioningMode::kForwardTraversal
: rocksdb::RangeDelPositioningMode::kFullScan;
std::vector<std::vector<rocksdb::PersistentRangeTombstone> >
all_persistent_range_tombstones(FLAGS_add_tombstones_per_run);
for (int i = 0; i < FLAGS_add_tombstones_per_run; i++) {
all_persistent_range_tombstones[i] =
std::vector<rocksdb::PersistentRangeTombstone>(
FLAGS_num_range_tombstones);
}
auto mode = FLAGS_use_collapsed
? rocksdb::RangeDelPositioningMode::kForwardTraversal
: rocksdb::RangeDelPositioningMode::kFullScan;
for (int i = 0; i < FLAGS_num_runs; i++) {
auto icmp = rocksdb::InternalKeyComparator(rocksdb::BytewiseComparator());
rocksdb::RangeDelAggregator range_del_agg(icmp, {} /* snapshots */,
FLAGS_use_collapsed);
// TODO(abhimadan): consider whether creating the range tombstones right
// before AddTombstones is artificially warming the cache compared to
// real workloads.
for (int j = 0; j < FLAGS_num_range_tombstones; j++) {
uint64_t start = rnd.Uniform(FLAGS_tombstone_start_upper_bound);
uint64_t end = start + std::max(1.0, normal_dist(random_gen));
persistent_range_tombstones[j] = rocksdb::PersistentRangeTombstone(
rocksdb::Key(start), rocksdb::Key(end), j);
}
for (auto& persistent_range_tombstones : all_persistent_range_tombstones) {
// TODO(abhimadan): consider whether creating the range tombstones right
// before AddTombstones is artificially warming the cache compared to
// real workloads.
for (int j = 0; j < FLAGS_num_range_tombstones; j++) {
uint64_t start = rnd.Uniform(FLAGS_tombstone_start_upper_bound);
uint64_t end = start + std::max(1.0, normal_dist(random_gen));
persistent_range_tombstones[j] = rocksdb::PersistentRangeTombstone(
rocksdb::Key(start), rocksdb::Key(end), j);
}
rocksdb::StopWatchNano stop_watch_add_tombstones(rocksdb::Env::Default(),
true /* auto_start */);
rocksdb::AddTombstones(&range_del_agg, persistent_range_tombstones);
stats.time_add_tombstones += stop_watch_add_tombstones.ElapsedNanos();
auto range_del_iter =
rocksdb::MakeRangeDelIterator(persistent_range_tombstones);
rocksdb::StopWatchNano stop_watch_add_tombstones(rocksdb::Env::Default(),
true /* auto_start */);
range_del_agg.AddTombstones(std::move(range_del_iter));
stats.time_add_tombstones += stop_watch_add_tombstones.ElapsedNanos();
}
rocksdb::ParsedInternalKey parsed_key;
parsed_key.sequence = FLAGS_num_range_tombstones / 2;