leveldb/benchmarks/db_bench_sqlite3.cc

723 lines
22 KiB
C++
Raw Normal View History

// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include <sqlite3.h>
#include <cstdio>
#include <cstdlib>
#include "util/histogram.h"
#include "util/random.h"
#include "util/testutil.h"
// Comma-separated list of operations to run in the specified order
// Actual benchmarks:
//
// fillseq -- write N values in sequential key order in async mode
// fillseqsync -- write N/100 values in sequential key order in sync mode
// fillseqbatch -- batch write N values in sequential key order in async mode
// fillrandom -- write N values in random key order in async mode
// fillrandsync -- write N/100 values in random key order in sync mode
// fillrandbatch -- batch write N values in sequential key order in async mode
// overwrite -- overwrite N values in random key order in async mode
// fillrand100K -- write N/1000 100K values in random order in async mode
// fillseq100K -- write N/1000 100K values in sequential order in async mode
// readseq -- read N times sequentially
// readrandom -- read N times in random order
// readrand100K -- read N/1000 100K values in sequential order in async mode
static const char* FLAGS_benchmarks =
"fillseq,"
"fillseqsync,"
"fillseqbatch,"
"fillrandom,"
"fillrandsync,"
"fillrandbatch,"
"overwrite,"
"overwritebatch,"
"readrandom,"
"readseq,"
"fillrand100K,"
"fillseq100K,"
"readseq,"
"readrand100K,";
// Number of key/values to place in database
static int FLAGS_num = 1000000;
// Number of read operations to do. If negative, do FLAGS_num reads.
static int FLAGS_reads = -1;
// Size of each value
static int FLAGS_value_size = 100;
// Print histogram of operation timings
static bool FLAGS_histogram = false;
// Arrange to generate values that shrink to this fraction of
// their original size after compression
static double FLAGS_compression_ratio = 0.5;
// Page size. Default 1 KB.
static int FLAGS_page_size = 1024;
// Number of pages.
// Default cache size = FLAGS_page_size * FLAGS_num_pages = 4 MB.
static int FLAGS_num_pages = 4096;
// If true, do not destroy the existing database. If you set this
// flag and also specify a benchmark that wants a fresh database, that
// benchmark will fail.
static bool FLAGS_use_existing_db = false;
Add WITHOUT ROWID to SQLite benchmark. The SQLite-specific schema feature is documented at https://www.sqlite.org/withoutrowid.html and https://www.sqlite.org/rowidtable.html. By default, SQLite stores each table in a B-tree keyed by an integer, called the ROWID. Any index, including the PRIMARY KEY index, is a separate B-tree mapping index keys to ROWIDs. Tables without ROWIDs are stored in a B-tree keyed by the primary key. Additional indexes (the PRIMARY KEY index is implicitly built into the table) are stored as B-trees mapping index keys to row primary keys. This CL introduces a boolean --use-rowids flag to db_bench_sqlite. When the flag is false (default), the schema of the test table includes WITHOUT ROWID. The test table uses a primary key, so adding WITHOUT ROWID to the schema reduces the number of B-trees used by the benchmark from 2 to 1. This brings SQLite's disk usage closer to LevelDB. When WITHOUT ROWID is used, SQLite fares better (than today) on benchmarks with small (16-byte) keys, and worse on benchmarks with large (100kb) keys. Baseline results: fillseq : 21.310 micros/op; 5.2 MB/s fillseqsync : 146.377 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.065 micros/op; 53.6 MB/s fillrandom : 34.767 micros/op; 3.2 MB/s fillrandsync : 159.943 micros/op; 0.7 MB/s (10000 ops) fillrandbatch : 15.055 micros/op; 7.3 MB/s overwrite : 43.660 micros/op; 2.5 MB/s overwritebatch : 27.691 micros/op; 4.0 MB/s readrandom : 12.725 micros/op; readseq : 2.602 micros/op; 36.7 MB/s fillrand100K : 606.333 micros/op; 157.3 MB/s (1000 ops) fillseq100K : 657.457 micros/op; 145.1 MB/s (1000 ops) readseq : 46.523 micros/op; 2049.9 MB/s readrand100K : 54.943 micros/op; Results after this CL: fillseq : 16.231 micros/op; 6.8 MB/s fillseqsync : 147.460 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.294 micros/op; 48.2 MB/s fillrandom : 27.871 micros/op; 4.0 MB/s fillrandsync : 141.979 micros/op; 0.8 MB/s (10000 ops) fillrandbatch : 16.087 micros/op; 6.9 MB/s overwrite : 26.829 micros/op; 4.1 MB/s overwritebatch : 19.014 micros/op; 5.8 MB/s readrandom : 11.657 micros/op; readseq : 0.155 micros/op; 615.0 MB/s fillrand100K : 816.812 micros/op; 116.8 MB/s (1000 ops) fillseq100K : 754.689 micros/op; 126.4 MB/s (1000 ops) readseq : 47.112 micros/op; 2024.3 MB/s readrand100K : 287.679 micros/op; Results after this CL, with --use-rowids=1 fillseq : 20.655 micros/op; 5.4 MB/s fillseqsync : 146.408 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.045 micros/op; 54.1 MB/s fillrandom : 34.080 micros/op; 3.2 MB/s fillrandsync : 154.582 micros/op; 0.7 MB/s (10000 ops) fillrandbatch : 14.404 micros/op; 7.7 MB/s overwrite : 42.928 micros/op; 2.6 MB/s overwritebatch : 27.829 micros/op; 4.0 MB/s readrandom : 12.835 micros/op; readseq : 2.483 micros/op; 38.4 MB/s fillrand100K : 603.265 micros/op; 158.1 MB/s (1000 ops) fillseq100K : 662.473 micros/op; 144.0 MB/s (1000 ops) readseq : 45.478 micros/op; 2097.0 MB/s readrand100K : 54.439 micros/op; PiperOrigin-RevId: 283407101
2019-12-03 05:37:34 +08:00
// If true, the SQLite table has ROWIDs.
static bool FLAGS_use_rowids = false;
// If true, we allow batch writes to occur
static bool FLAGS_transaction = true;
// If true, we enable Write-Ahead Logging
static bool FLAGS_WAL_enabled = true;
// Use the db with the following name.
static const char* FLAGS_db = nullptr;
inline static void ExecErrorCheck(int status, char* err_msg) {
if (status != SQLITE_OK) {
fprintf(stderr, "SQL error: %s\n", err_msg);
sqlite3_free(err_msg);
exit(1);
}
}
inline static void StepErrorCheck(int status) {
if (status != SQLITE_DONE) {
fprintf(stderr, "SQL step error: status = %d\n", status);
exit(1);
}
}
inline static void ErrorCheck(int status) {
if (status != SQLITE_OK) {
fprintf(stderr, "sqlite3 error: status = %d\n", status);
exit(1);
}
}
inline static void WalCheckpoint(sqlite3* db_) {
// Flush all writes to disk
if (FLAGS_WAL_enabled) {
sqlite3_wal_checkpoint_v2(db_, nullptr, SQLITE_CHECKPOINT_FULL, nullptr,
nullptr);
}
}
namespace leveldb {
// Helper for quickly generating random data.
namespace {
class RandomGenerator {
private:
std::string data_;
int pos_;
public:
RandomGenerator() {
// We use a limited amount of data over and over again and ensure
// that it is larger than the compression window (32KB), and also
// large enough to serve all typical value sizes we want to write.
Random rnd(301);
std::string piece;
while (data_.size() < 1048576) {
// Add a short fragment that is as compressible as specified
// by FLAGS_compression_ratio.
test::CompressibleString(&rnd, FLAGS_compression_ratio, 100, &piece);
data_.append(piece);
}
pos_ = 0;
}
Slice Generate(int len) {
if (pos_ + len > data_.size()) {
pos_ = 0;
assert(len < data_.size());
}
pos_ += len;
return Slice(data_.data() + pos_ - len, len);
}
};
static Slice TrimSpace(Slice s) {
int start = 0;
while (start < s.size() && isspace(s[start])) {
start++;
}
int limit = s.size();
while (limit > start && isspace(s[limit - 1])) {
limit--;
}
return Slice(s.data() + start, limit - start);
}
} // namespace
class Benchmark {
private:
sqlite3* db_;
int db_num_;
int num_;
int reads_;
double start_;
double last_op_finish_;
int64_t bytes_;
std::string message_;
Histogram hist_;
RandomGenerator gen_;
Random rand_;
// State kept for progress messages
int done_;
int next_report_; // When to report next
void PrintHeader() {
const int kKeySize = 16;
PrintEnvironment();
fprintf(stdout, "Keys: %d bytes each\n", kKeySize);
fprintf(stdout, "Values: %d bytes each\n", FLAGS_value_size);
fprintf(stdout, "Entries: %d\n", num_);
fprintf(stdout, "RawSize: %.1f MB (estimated)\n",
((static_cast<int64_t>(kKeySize + FLAGS_value_size) * num_) /
1048576.0));
PrintWarnings();
fprintf(stdout, "------------------------------------------------\n");
}
void PrintWarnings() {
#if defined(__GNUC__) && !defined(__OPTIMIZE__)
fprintf(
stdout,
"WARNING: Optimization is disabled: benchmarks unnecessarily slow\n");
#endif
#ifndef NDEBUG
fprintf(stdout,
"WARNING: Assertions are enabled; benchmarks unnecessarily slow\n");
#endif
}
void PrintEnvironment() {
fprintf(stderr, "SQLite: version %s\n", SQLITE_VERSION);
#if defined(__linux)
time_t now = time(nullptr);
fprintf(stderr, "Date: %s", ctime(&now)); // ctime() adds newline
FILE* cpuinfo = fopen("/proc/cpuinfo", "r");
if (cpuinfo != nullptr) {
char line[1000];
int num_cpus = 0;
std::string cpu_type;
std::string cache_size;
while (fgets(line, sizeof(line), cpuinfo) != nullptr) {
const char* sep = strchr(line, ':');
if (sep == nullptr) {
continue;
}
Slice key = TrimSpace(Slice(line, sep - 1 - line));
Slice val = TrimSpace(Slice(sep + 1));
if (key == "model name") {
++num_cpus;
cpu_type = val.ToString();
} else if (key == "cache size") {
cache_size = val.ToString();
}
}
fclose(cpuinfo);
fprintf(stderr, "CPU: %d * %s\n", num_cpus, cpu_type.c_str());
fprintf(stderr, "CPUCache: %s\n", cache_size.c_str());
}
#endif
}
void Start() {
start_ = Env::Default()->NowMicros() * 1e-6;
bytes_ = 0;
message_.clear();
last_op_finish_ = start_;
hist_.Clear();
done_ = 0;
next_report_ = 100;
}
void FinishedSingleOp() {
if (FLAGS_histogram) {
double now = Env::Default()->NowMicros() * 1e-6;
double micros = (now - last_op_finish_) * 1e6;
hist_.Add(micros);
if (micros > 20000) {
fprintf(stderr, "long op: %.1f micros%30s\r", micros, "");
fflush(stderr);
}
last_op_finish_ = now;
}
done_++;
if (done_ >= next_report_) {
if (next_report_ < 1000)
next_report_ += 100;
else if (next_report_ < 5000)
next_report_ += 500;
else if (next_report_ < 10000)
next_report_ += 1000;
else if (next_report_ < 50000)
next_report_ += 5000;
else if (next_report_ < 100000)
next_report_ += 10000;
else if (next_report_ < 500000)
next_report_ += 50000;
else
next_report_ += 100000;
fprintf(stderr, "... finished %d ops%30s\r", done_, "");
fflush(stderr);
}
}
void Stop(const Slice& name) {
double finish = Env::Default()->NowMicros() * 1e-6;
// Pretend at least one op was done in case we are running a benchmark
// that does not call FinishedSingleOp().
if (done_ < 1) done_ = 1;
if (bytes_ > 0) {
char rate[100];
snprintf(rate, sizeof(rate), "%6.1f MB/s",
(bytes_ / 1048576.0) / (finish - start_));
if (!message_.empty()) {
message_ = std::string(rate) + " " + message_;
} else {
message_ = rate;
}
}
fprintf(stdout, "%-12s : %11.3f micros/op;%s%s\n", name.ToString().c_str(),
(finish - start_) * 1e6 / done_, (message_.empty() ? "" : " "),
message_.c_str());
if (FLAGS_histogram) {
fprintf(stdout, "Microseconds per op:\n%s\n", hist_.ToString().c_str());
}
fflush(stdout);
}
public:
enum Order { SEQUENTIAL, RANDOM };
enum DBState { FRESH, EXISTING };
Benchmark()
: db_(nullptr),
db_num_(0),
num_(FLAGS_num),
reads_(FLAGS_reads < 0 ? FLAGS_num : FLAGS_reads),
bytes_(0),
rand_(301) {
std::vector<std::string> files;
std::string test_dir;
Env::Default()->GetTestDirectory(&test_dir);
Env::Default()->GetChildren(test_dir, &files);
if (!FLAGS_use_existing_db) {
for (int i = 0; i < files.size(); i++) {
if (Slice(files[i]).starts_with("dbbench_sqlite3")) {
std::string file_name(test_dir);
file_name += "/";
file_name += files[i];
Add Env::Remove{File,Dir} which obsolete Env::Delete{File,Dir}. The "DeleteFile" method name causes pain for Windows developers, because <windows.h> #defines a DeleteFile macro to DeleteFileW or DeleteFileA. Current code uses workarounds, like #undefining DeleteFile everywhere an Env is declared, implemented, or used. This CL removes the need for workarounds by renaming Env::DeleteFile to Env::RemoveFile. For consistency, Env::DeleteDir is also renamed to Env::RemoveDir. A few internal methods are also renamed for consistency. Software that supports Windows is expected to migrate any Env implementations and usage to Remove{File,Dir}, and never use the name Env::Delete{File,Dir} in its code. The renaming is done in a backwards-compatible way, at the risk of making it slightly more difficult to build a new correct Env implementation. The backwards compatibility is achieved using the following hacks: 1) Env::Remove{File,Dir} methods are added, with a default implementation that calls into Env::Delete{File,Dir}. This makes old Env implementations compatible with code that calls into the updated API. 2) The Env::Delete{File,Dir} methods are no longer pure virtuals. Instead, they gain a default implementation that calls into Env::Remove{File,Dir}. This makes updated Env implementations compatible with code that calls into the old API. The cost of this approach is that it's possible to write an Env without overriding either Rename{File,Dir} or Delete{File,Dir}, without getting a compiler warning. However, attempting to run the test suite will immediately fail with an infinite call stack ending in {Remove,Delete}{File,Dir}, making developers aware of the problem. PiperOrigin-RevId: 288710907
2020-01-09 01:14:53 +08:00
Env::Default()->RemoveFile(file_name.c_str());
}
}
}
}
~Benchmark() {
int status = sqlite3_close(db_);
ErrorCheck(status);
}
void Run() {
PrintHeader();
Open();
const char* benchmarks = FLAGS_benchmarks;
while (benchmarks != nullptr) {
const char* sep = strchr(benchmarks, ',');
Slice name;
if (sep == nullptr) {
name = benchmarks;
benchmarks = nullptr;
} else {
name = Slice(benchmarks, sep - benchmarks);
benchmarks = sep + 1;
}
bytes_ = 0;
Start();
bool known = true;
bool write_sync = false;
if (name == Slice("fillseq")) {
Write(write_sync, SEQUENTIAL, FRESH, num_, FLAGS_value_size, 1);
WalCheckpoint(db_);
} else if (name == Slice("fillseqbatch")) {
Write(write_sync, SEQUENTIAL, FRESH, num_, FLAGS_value_size, 1000);
WalCheckpoint(db_);
} else if (name == Slice("fillrandom")) {
Write(write_sync, RANDOM, FRESH, num_, FLAGS_value_size, 1);
WalCheckpoint(db_);
} else if (name == Slice("fillrandbatch")) {
Write(write_sync, RANDOM, FRESH, num_, FLAGS_value_size, 1000);
WalCheckpoint(db_);
} else if (name == Slice("overwrite")) {
Write(write_sync, RANDOM, EXISTING, num_, FLAGS_value_size, 1);
WalCheckpoint(db_);
} else if (name == Slice("overwritebatch")) {
Write(write_sync, RANDOM, EXISTING, num_, FLAGS_value_size, 1000);
WalCheckpoint(db_);
} else if (name == Slice("fillrandsync")) {
write_sync = true;
Write(write_sync, RANDOM, FRESH, num_ / 100, FLAGS_value_size, 1);
WalCheckpoint(db_);
} else if (name == Slice("fillseqsync")) {
write_sync = true;
Write(write_sync, SEQUENTIAL, FRESH, num_ / 100, FLAGS_value_size, 1);
WalCheckpoint(db_);
} else if (name == Slice("fillrand100K")) {
Write(write_sync, RANDOM, FRESH, num_ / 1000, 100 * 1000, 1);
WalCheckpoint(db_);
} else if (name == Slice("fillseq100K")) {
Write(write_sync, SEQUENTIAL, FRESH, num_ / 1000, 100 * 1000, 1);
WalCheckpoint(db_);
} else if (name == Slice("readseq")) {
ReadSequential();
} else if (name == Slice("readrandom")) {
Read(RANDOM, 1);
} else if (name == Slice("readrand100K")) {
int n = reads_;
reads_ /= 1000;
Read(RANDOM, 1);
reads_ = n;
} else {
known = false;
if (name != Slice()) { // No error message for empty name
fprintf(stderr, "unknown benchmark '%s'\n", name.ToString().c_str());
}
}
if (known) {
Stop(name);
}
}
}
void Open() {
assert(db_ == nullptr);
int status;
char file_name[100];
char* err_msg = nullptr;
db_num_++;
// Open database
std::string tmp_dir;
Env::Default()->GetTestDirectory(&tmp_dir);
snprintf(file_name, sizeof(file_name), "%s/dbbench_sqlite3-%d.db",
tmp_dir.c_str(), db_num_);
status = sqlite3_open(file_name, &db_);
if (status) {
fprintf(stderr, "open error: %s\n", sqlite3_errmsg(db_));
exit(1);
}
// Change SQLite cache size
char cache_size[100];
snprintf(cache_size, sizeof(cache_size), "PRAGMA cache_size = %d",
FLAGS_num_pages);
status = sqlite3_exec(db_, cache_size, nullptr, nullptr, &err_msg);
ExecErrorCheck(status, err_msg);
// FLAGS_page_size is defaulted to 1024
if (FLAGS_page_size != 1024) {
char page_size[100];
snprintf(page_size, sizeof(page_size), "PRAGMA page_size = %d",
FLAGS_page_size);
status = sqlite3_exec(db_, page_size, nullptr, nullptr, &err_msg);
ExecErrorCheck(status, err_msg);
}
// Change journal mode to WAL if WAL enabled flag is on
if (FLAGS_WAL_enabled) {
std::string WAL_stmt = "PRAGMA journal_mode = WAL";
// LevelDB's default cache size is a combined 4 MB
std::string WAL_checkpoint = "PRAGMA wal_autocheckpoint = 4096";
status = sqlite3_exec(db_, WAL_stmt.c_str(), nullptr, nullptr, &err_msg);
ExecErrorCheck(status, err_msg);
status =
sqlite3_exec(db_, WAL_checkpoint.c_str(), nullptr, nullptr, &err_msg);
ExecErrorCheck(status, err_msg);
}
// Change locking mode to exclusive and create tables/index for database
std::string locking_stmt = "PRAGMA locking_mode = EXCLUSIVE";
std::string create_stmt =
"CREATE TABLE test (key blob, value blob, PRIMARY KEY(key))";
Add WITHOUT ROWID to SQLite benchmark. The SQLite-specific schema feature is documented at https://www.sqlite.org/withoutrowid.html and https://www.sqlite.org/rowidtable.html. By default, SQLite stores each table in a B-tree keyed by an integer, called the ROWID. Any index, including the PRIMARY KEY index, is a separate B-tree mapping index keys to ROWIDs. Tables without ROWIDs are stored in a B-tree keyed by the primary key. Additional indexes (the PRIMARY KEY index is implicitly built into the table) are stored as B-trees mapping index keys to row primary keys. This CL introduces a boolean --use-rowids flag to db_bench_sqlite. When the flag is false (default), the schema of the test table includes WITHOUT ROWID. The test table uses a primary key, so adding WITHOUT ROWID to the schema reduces the number of B-trees used by the benchmark from 2 to 1. This brings SQLite's disk usage closer to LevelDB. When WITHOUT ROWID is used, SQLite fares better (than today) on benchmarks with small (16-byte) keys, and worse on benchmarks with large (100kb) keys. Baseline results: fillseq : 21.310 micros/op; 5.2 MB/s fillseqsync : 146.377 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.065 micros/op; 53.6 MB/s fillrandom : 34.767 micros/op; 3.2 MB/s fillrandsync : 159.943 micros/op; 0.7 MB/s (10000 ops) fillrandbatch : 15.055 micros/op; 7.3 MB/s overwrite : 43.660 micros/op; 2.5 MB/s overwritebatch : 27.691 micros/op; 4.0 MB/s readrandom : 12.725 micros/op; readseq : 2.602 micros/op; 36.7 MB/s fillrand100K : 606.333 micros/op; 157.3 MB/s (1000 ops) fillseq100K : 657.457 micros/op; 145.1 MB/s (1000 ops) readseq : 46.523 micros/op; 2049.9 MB/s readrand100K : 54.943 micros/op; Results after this CL: fillseq : 16.231 micros/op; 6.8 MB/s fillseqsync : 147.460 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.294 micros/op; 48.2 MB/s fillrandom : 27.871 micros/op; 4.0 MB/s fillrandsync : 141.979 micros/op; 0.8 MB/s (10000 ops) fillrandbatch : 16.087 micros/op; 6.9 MB/s overwrite : 26.829 micros/op; 4.1 MB/s overwritebatch : 19.014 micros/op; 5.8 MB/s readrandom : 11.657 micros/op; readseq : 0.155 micros/op; 615.0 MB/s fillrand100K : 816.812 micros/op; 116.8 MB/s (1000 ops) fillseq100K : 754.689 micros/op; 126.4 MB/s (1000 ops) readseq : 47.112 micros/op; 2024.3 MB/s readrand100K : 287.679 micros/op; Results after this CL, with --use-rowids=1 fillseq : 20.655 micros/op; 5.4 MB/s fillseqsync : 146.408 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.045 micros/op; 54.1 MB/s fillrandom : 34.080 micros/op; 3.2 MB/s fillrandsync : 154.582 micros/op; 0.7 MB/s (10000 ops) fillrandbatch : 14.404 micros/op; 7.7 MB/s overwrite : 42.928 micros/op; 2.6 MB/s overwritebatch : 27.829 micros/op; 4.0 MB/s readrandom : 12.835 micros/op; readseq : 2.483 micros/op; 38.4 MB/s fillrand100K : 603.265 micros/op; 158.1 MB/s (1000 ops) fillseq100K : 662.473 micros/op; 144.0 MB/s (1000 ops) readseq : 45.478 micros/op; 2097.0 MB/s readrand100K : 54.439 micros/op; PiperOrigin-RevId: 283407101
2019-12-03 05:37:34 +08:00
if (!FLAGS_use_rowids) create_stmt += " WITHOUT ROWID";
std::string stmt_array[] = {locking_stmt, create_stmt};
int stmt_array_length = sizeof(stmt_array) / sizeof(std::string);
for (int i = 0; i < stmt_array_length; i++) {
status =
sqlite3_exec(db_, stmt_array[i].c_str(), nullptr, nullptr, &err_msg);
ExecErrorCheck(status, err_msg);
}
}
void Write(bool write_sync, Order order, DBState state, int num_entries,
int value_size, int entries_per_batch) {
// Create new database if state == FRESH
if (state == FRESH) {
if (FLAGS_use_existing_db) {
message_ = "skipping (--use_existing_db is true)";
return;
}
sqlite3_close(db_);
db_ = nullptr;
Open();
Start();
}
if (num_entries != num_) {
char msg[100];
snprintf(msg, sizeof(msg), "(%d ops)", num_entries);
message_ = msg;
}
char* err_msg = nullptr;
int status;
sqlite3_stmt *replace_stmt, *begin_trans_stmt, *end_trans_stmt;
std::string replace_str = "REPLACE INTO test (key, value) VALUES (?, ?)";
std::string begin_trans_str = "BEGIN TRANSACTION;";
std::string end_trans_str = "END TRANSACTION;";
// Check for synchronous flag in options
std::string sync_stmt =
(write_sync) ? "PRAGMA synchronous = FULL" : "PRAGMA synchronous = OFF";
status = sqlite3_exec(db_, sync_stmt.c_str(), nullptr, nullptr, &err_msg);
ExecErrorCheck(status, err_msg);
// Preparing sqlite3 statements
status = sqlite3_prepare_v2(db_, replace_str.c_str(), -1, &replace_stmt,
nullptr);
ErrorCheck(status);
status = sqlite3_prepare_v2(db_, begin_trans_str.c_str(), -1,
&begin_trans_stmt, nullptr);
ErrorCheck(status);
status = sqlite3_prepare_v2(db_, end_trans_str.c_str(), -1, &end_trans_stmt,
nullptr);
ErrorCheck(status);
bool transaction = (entries_per_batch > 1);
for (int i = 0; i < num_entries; i += entries_per_batch) {
// Begin write transaction
if (FLAGS_transaction && transaction) {
status = sqlite3_step(begin_trans_stmt);
StepErrorCheck(status);
status = sqlite3_reset(begin_trans_stmt);
ErrorCheck(status);
}
// Create and execute SQL statements
for (int j = 0; j < entries_per_batch; j++) {
const char* value = gen_.Generate(value_size).data();
// Create values for key-value pair
const int k =
(order == SEQUENTIAL) ? i + j : (rand_.Next() % num_entries);
char key[100];
snprintf(key, sizeof(key), "%016d", k);
// Bind KV values into replace_stmt
status = sqlite3_bind_blob(replace_stmt, 1, key, 16, SQLITE_STATIC);
ErrorCheck(status);
status = sqlite3_bind_blob(replace_stmt, 2, value, value_size,
SQLITE_STATIC);
ErrorCheck(status);
// Execute replace_stmt
bytes_ += value_size + strlen(key);
status = sqlite3_step(replace_stmt);
StepErrorCheck(status);
// Reset SQLite statement for another use
status = sqlite3_clear_bindings(replace_stmt);
ErrorCheck(status);
status = sqlite3_reset(replace_stmt);
ErrorCheck(status);
FinishedSingleOp();
}
// End write transaction
if (FLAGS_transaction && transaction) {
status = sqlite3_step(end_trans_stmt);
StepErrorCheck(status);
status = sqlite3_reset(end_trans_stmt);
ErrorCheck(status);
}
}
status = sqlite3_finalize(replace_stmt);
ErrorCheck(status);
status = sqlite3_finalize(begin_trans_stmt);
ErrorCheck(status);
status = sqlite3_finalize(end_trans_stmt);
ErrorCheck(status);
}
void Read(Order order, int entries_per_batch) {
int status;
sqlite3_stmt *read_stmt, *begin_trans_stmt, *end_trans_stmt;
std::string read_str = "SELECT * FROM test WHERE key = ?";
std::string begin_trans_str = "BEGIN TRANSACTION;";
std::string end_trans_str = "END TRANSACTION;";
// Preparing sqlite3 statements
status = sqlite3_prepare_v2(db_, begin_trans_str.c_str(), -1,
&begin_trans_stmt, nullptr);
ErrorCheck(status);
status = sqlite3_prepare_v2(db_, end_trans_str.c_str(), -1, &end_trans_stmt,
nullptr);
ErrorCheck(status);
status = sqlite3_prepare_v2(db_, read_str.c_str(), -1, &read_stmt, nullptr);
ErrorCheck(status);
bool transaction = (entries_per_batch > 1);
for (int i = 0; i < reads_; i += entries_per_batch) {
// Begin read transaction
if (FLAGS_transaction && transaction) {
status = sqlite3_step(begin_trans_stmt);
StepErrorCheck(status);
status = sqlite3_reset(begin_trans_stmt);
ErrorCheck(status);
}
// Create and execute SQL statements
for (int j = 0; j < entries_per_batch; j++) {
// Create key value
char key[100];
int k = (order == SEQUENTIAL) ? i + j : (rand_.Next() % reads_);
snprintf(key, sizeof(key), "%016d", k);
// Bind key value into read_stmt
status = sqlite3_bind_blob(read_stmt, 1, key, 16, SQLITE_STATIC);
ErrorCheck(status);
// Execute read statement
while ((status = sqlite3_step(read_stmt)) == SQLITE_ROW) {
}
StepErrorCheck(status);
// Reset SQLite statement for another use
status = sqlite3_clear_bindings(read_stmt);
ErrorCheck(status);
status = sqlite3_reset(read_stmt);
ErrorCheck(status);
FinishedSingleOp();
}
// End read transaction
if (FLAGS_transaction && transaction) {
status = sqlite3_step(end_trans_stmt);
StepErrorCheck(status);
status = sqlite3_reset(end_trans_stmt);
ErrorCheck(status);
}
}
status = sqlite3_finalize(read_stmt);
ErrorCheck(status);
status = sqlite3_finalize(begin_trans_stmt);
ErrorCheck(status);
status = sqlite3_finalize(end_trans_stmt);
ErrorCheck(status);
}
void ReadSequential() {
int status;
sqlite3_stmt* pStmt;
std::string read_str = "SELECT * FROM test ORDER BY key";
status = sqlite3_prepare_v2(db_, read_str.c_str(), -1, &pStmt, nullptr);
ErrorCheck(status);
for (int i = 0; i < reads_ && SQLITE_ROW == sqlite3_step(pStmt); i++) {
bytes_ += sqlite3_column_bytes(pStmt, 1) + sqlite3_column_bytes(pStmt, 2);
FinishedSingleOp();
}
status = sqlite3_finalize(pStmt);
ErrorCheck(status);
}
};
} // namespace leveldb
int main(int argc, char** argv) {
std::string default_db_path;
for (int i = 1; i < argc; i++) {
double d;
int n;
char junk;
if (leveldb::Slice(argv[i]).starts_with("--benchmarks=")) {
FLAGS_benchmarks = argv[i] + strlen("--benchmarks=");
} else if (sscanf(argv[i], "--histogram=%d%c", &n, &junk) == 1 &&
(n == 0 || n == 1)) {
FLAGS_histogram = n;
} else if (sscanf(argv[i], "--compression_ratio=%lf%c", &d, &junk) == 1) {
FLAGS_compression_ratio = d;
} else if (sscanf(argv[i], "--use_existing_db=%d%c", &n, &junk) == 1 &&
(n == 0 || n == 1)) {
FLAGS_use_existing_db = n;
Add WITHOUT ROWID to SQLite benchmark. The SQLite-specific schema feature is documented at https://www.sqlite.org/withoutrowid.html and https://www.sqlite.org/rowidtable.html. By default, SQLite stores each table in a B-tree keyed by an integer, called the ROWID. Any index, including the PRIMARY KEY index, is a separate B-tree mapping index keys to ROWIDs. Tables without ROWIDs are stored in a B-tree keyed by the primary key. Additional indexes (the PRIMARY KEY index is implicitly built into the table) are stored as B-trees mapping index keys to row primary keys. This CL introduces a boolean --use-rowids flag to db_bench_sqlite. When the flag is false (default), the schema of the test table includes WITHOUT ROWID. The test table uses a primary key, so adding WITHOUT ROWID to the schema reduces the number of B-trees used by the benchmark from 2 to 1. This brings SQLite's disk usage closer to LevelDB. When WITHOUT ROWID is used, SQLite fares better (than today) on benchmarks with small (16-byte) keys, and worse on benchmarks with large (100kb) keys. Baseline results: fillseq : 21.310 micros/op; 5.2 MB/s fillseqsync : 146.377 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.065 micros/op; 53.6 MB/s fillrandom : 34.767 micros/op; 3.2 MB/s fillrandsync : 159.943 micros/op; 0.7 MB/s (10000 ops) fillrandbatch : 15.055 micros/op; 7.3 MB/s overwrite : 43.660 micros/op; 2.5 MB/s overwritebatch : 27.691 micros/op; 4.0 MB/s readrandom : 12.725 micros/op; readseq : 2.602 micros/op; 36.7 MB/s fillrand100K : 606.333 micros/op; 157.3 MB/s (1000 ops) fillseq100K : 657.457 micros/op; 145.1 MB/s (1000 ops) readseq : 46.523 micros/op; 2049.9 MB/s readrand100K : 54.943 micros/op; Results after this CL: fillseq : 16.231 micros/op; 6.8 MB/s fillseqsync : 147.460 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.294 micros/op; 48.2 MB/s fillrandom : 27.871 micros/op; 4.0 MB/s fillrandsync : 141.979 micros/op; 0.8 MB/s (10000 ops) fillrandbatch : 16.087 micros/op; 6.9 MB/s overwrite : 26.829 micros/op; 4.1 MB/s overwritebatch : 19.014 micros/op; 5.8 MB/s readrandom : 11.657 micros/op; readseq : 0.155 micros/op; 615.0 MB/s fillrand100K : 816.812 micros/op; 116.8 MB/s (1000 ops) fillseq100K : 754.689 micros/op; 126.4 MB/s (1000 ops) readseq : 47.112 micros/op; 2024.3 MB/s readrand100K : 287.679 micros/op; Results after this CL, with --use-rowids=1 fillseq : 20.655 micros/op; 5.4 MB/s fillseqsync : 146.408 micros/op; 0.8 MB/s (10000 ops) fillseqbatch : 2.045 micros/op; 54.1 MB/s fillrandom : 34.080 micros/op; 3.2 MB/s fillrandsync : 154.582 micros/op; 0.7 MB/s (10000 ops) fillrandbatch : 14.404 micros/op; 7.7 MB/s overwrite : 42.928 micros/op; 2.6 MB/s overwritebatch : 27.829 micros/op; 4.0 MB/s readrandom : 12.835 micros/op; readseq : 2.483 micros/op; 38.4 MB/s fillrand100K : 603.265 micros/op; 158.1 MB/s (1000 ops) fillseq100K : 662.473 micros/op; 144.0 MB/s (1000 ops) readseq : 45.478 micros/op; 2097.0 MB/s readrand100K : 54.439 micros/op; PiperOrigin-RevId: 283407101
2019-12-03 05:37:34 +08:00
} else if (sscanf(argv[i], "--use_rowids=%d%c", &n, &junk) == 1 &&
(n == 0 || n == 1)) {
FLAGS_use_rowids = n;
} else if (sscanf(argv[i], "--num=%d%c", &n, &junk) == 1) {
FLAGS_num = n;
} else if (sscanf(argv[i], "--reads=%d%c", &n, &junk) == 1) {
FLAGS_reads = n;
} else if (sscanf(argv[i], "--value_size=%d%c", &n, &junk) == 1) {
FLAGS_value_size = n;
} else if (leveldb::Slice(argv[i]) == leveldb::Slice("--no_transaction")) {
FLAGS_transaction = false;
} else if (sscanf(argv[i], "--page_size=%d%c", &n, &junk) == 1) {
FLAGS_page_size = n;
} else if (sscanf(argv[i], "--num_pages=%d%c", &n, &junk) == 1) {
FLAGS_num_pages = n;
} else if (sscanf(argv[i], "--WAL_enabled=%d%c", &n, &junk) == 1 &&
(n == 0 || n == 1)) {
FLAGS_WAL_enabled = n;
} else if (strncmp(argv[i], "--db=", 5) == 0) {
FLAGS_db = argv[i] + 5;
} else {
fprintf(stderr, "Invalid flag '%s'\n", argv[i]);
exit(1);
}
}
// Choose a location for the test database if none given with --db=<path>
if (FLAGS_db == nullptr) {
leveldb::Env::Default()->GetTestDirectory(&default_db_path);
default_db_path += "/dbbench";
FLAGS_db = default_db_path.c_str();
}
leveldb::Benchmark benchmark;
benchmark.Run();
return 0;
}