# Copyright 2018-2019 Florian Fischer # # This file is part of allocbench. # # allocbench is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # allocbench is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with allocbench. If not, see . """sysbench SQL read-only benchmark This benchmark is heavily inspired by a blog post from Alexey Stroganov from Percona: https://web.archive.org/web/20190706104404/https://www.percona.com/blog/2012/07/05/impact-of-memory-allocators-on-mysql-performance/ It uses the read-only database benchmark from sysbench, a commonly used system benchmarking tool to measure the performance of mysqld per allocator. The read-only benchmark on a relatively small database (~1GB) is used to omit I/O latency and be cpu-bound, maximizing the allocators influence on performance. Behavior per allocator: * Start mysqld using the allocator * Run sysbench oltp_read_only once per thread count * Shutdown mysqld 40 Thread workload: * allocator calls: 1519226 * malloc: 722421 (47.55%) * free: 795231 (52.34%) * calloc: 1501 (0.10%) * realloc: 73 (0.004%) * Approximate allocator ratios: * malloc: 0.69% * free: 0.36% * calloc: 0.04% * Top 10 allocation sizes 71.36% of all allocations 1. 288 B occurred 112556 times 2. 4064 B occurred 112552 times 3. 9 B occurred 61978 times 4. 328 B occurred 56275 times 5. 64 B occurred 48498 times 6. 1040 B occurred 28174 times 7. 360 B occurred 28140 times 8. 65544 B occurred 28136 times 9. 104 B occurred 25794 times 10. 992 B occurred 14521 times allocations <= 64: 131723 18.19% allocations <= 1024: 423315 58.47% allocations <= 4096: 622732 86.01% mysqld starts one thread per connection, which produce roughly the same allocator workload (checked with a malt trace). Interpretation: The mysql benchmark tries to be as near as possible to a real world workload. So all non-functional characteristics of an allocator are measured. This means the results can give hints on how each allocator performs for a similar workload. But the results don't directly explain why an allocator performs this way. To obtain a more complete understanding deeper analysis of the allocators algorithm, host system and workload is needed. """ import multiprocessing import os import re import shutil import subprocess from subprocess import PIPE import sys import numpy as np from src.benchmark import Benchmark import src.facter from src.util import print_status, print_debug, print_info2 MYSQL_USER = "root" RUN_TIME = 300 TABLES = 5 PREPARE_CMD = (f"sysbench oltp_read_only --db-driver=mysql --mysql-user={MYSQL_USER} " f"--threads={multiprocessing.cpu_count()} " f"--mysql-socket={{build_dir}}/socket --tables={TABLES} --table-size=1000000 prepare") CMD = (f"sysbench oltp_read_only --threads={{nthreads}} --time={RUN_TIME} --tables={TABLES} " f"--db-driver=mysql --mysql-user={MYSQL_USER} --mysql-socket={{build_dir}}/socket run") SERVER_CMD = ("mysqld --no-defaults -h {build_dir} --socket={build_dir}/socket --port=123456 " f"--max-connections={multiprocessing.cpu_count()} --secure-file-priv=") class BenchmarkMYSQL(Benchmark): """Mysql bechmark definition""" def __init__(self): name = "mysql" self.args = {"nthreads": Benchmark.scale_threads_for_cpus(1)} self.cmd = CMD self.servers = [{"name": "mysqld", "cmd" : SERVER_CMD}] self.measure_cmd = "" self.requirements = ["mysqld", "sysbench"] super().__init__(name) self.results["facts"]["runtime [s]"] = RUN_TIME def prepare(self): super().prepare() # save mysqld and sysbench versions for exe in self.requirements: self.results["facts"]["versions"][exe] = src.facter.exe_version(exe, "--version") # Setup Test Environment if not os.path.exists(self.build_dir): print_status("Prepare mysqld directory and database") os.makedirs(self.build_dir) # Init database if "MariaDB" in self.results["facts"]["versions"]["mysqld"]: init_db_cmd = ["mysql_install_db", "--basedir=/usr", f"--datadir={self.build_dir}"] print_info2("MariaDB detected") else: init_db_cmd = ["mysqld", "-h", self.build_dir, "--initialize-insecure"] print_info2("Oracle MySQL detected") p = subprocess.run(init_db_cmd, stdout=PIPE, stderr=PIPE) if not p.returncode == 0: print_debug(init_db_cmd) print_debug("Stdout:", p.stdout, file=sys.stdout) print_debug("Stderr:", p.stderr, file=sys.stderr) raise Exception("Creating test DB failed with:", p.returncode) self.start_servers() # Create sbtest TABLE p = subprocess.run(f"mysql -u {MYSQL_USER} -S {self.build_dir}/socket".split(), input=b"CREATE DATABASE sbtest;\n", stdout=PIPE, stderr=PIPE, cwd=self.build_dir) if p.returncode != 0: print_debug("Stderr:", p.stderr, file=sys.stderr) raise Exception("Creating test tables failed with:", p.returncode) print_status("Prepare test tables ...") prepare_cmd = PREPARE_CMD.format(build_dir=self.build_dir) p = subprocess.run(prepare_cmd.split(), stdout=PIPE, stderr=PIPE) if p.returncode != 0: print_debug(f"Cmd: {prepare_cmd} failed with {p.returncode}", file=sys.stderr) print_debug("Stdout:", p.stdout, file=sys.stderr) print_debug("Stderr:", p.stderr, file=sys.stderr) raise Exception("Preparing test tables failed with:", p.returncode) self.shutdown_servers() def process_output(self, result, stdout, stderr, allocator, perm): result["transactions"] = re.search("transactions:\\s*(\\d*)", stdout).group(1) result["queries"] = re.search("queries:\\s*(\\d*)", stdout).group(1) # Latency result["min"] = re.search("min:\\s*(\\d*.\\d*)", stdout).group(1) result["avg"] = re.search("avg:\\s*(\\d*.\\d*)", stdout).group(1) result["max"] = re.search("max:\\s*(\\d*.\\d*)", stdout).group(1) with open("/proc/"+str(self.servers[0]["popen"].pid)+"/status", "r") as f: for l in f.readlines(): if l.startswith("VmHWM:"): result["rssmax"] = int(l.replace("VmHWM:", "").strip().split()[0]) break def summary(self): allocators = self.results["allocators"] args = self.results["args"] # linear plot self.plot_single_arg("{transactions}", xlabel='"threads"', ylabel='"transactions"', title='"sysbench oltp read only"', filepostfix="l") # normalized linear plot ref_alloc = list(allocators)[0] self.plot_single_arg("{transactions}", xlabel='"threads"', ylabel='"transactions scaled at " + scale', title='"sysbench oltp read only"', filepostfix="norm.l", scale=ref_alloc) # bar plot self.barplot_single_arg("{transactions}", xlabel='"threads"', ylabel='"transactions"', title='"sysbench oltp read only"', filepostfix="b") # normalized bar plot self.barplot_single_arg("{transactions}", xlabel='"threads"', ylabel='"transactions scaled at " + scale', title='"sysbench oltp read only"', filepostfix="norm.b", scale=ref_alloc) # Memusage self.barplot_single_arg("{rssmax}", xlabel='"threads"', ylabel='"VmHWM in kB"', title='"Memusage sysbench oltp read only"', filepostfix="mem") # Colored latex table showing transactions count d = {allocator: {} for allocator in allocators} for perm in self.iterate_args(args=args): for allocator in allocators: transactions = [float(measure["transactions"]) for measure in self.results[allocator][perm]] mean = np.mean(transactions) std = np.std(transactions)/mean d[allocator][perm] = {"mean": mean, "std": std} mins = {} maxs = {} for perm in self.iterate_args(args=args): cmax = None cmin = None for i, allocator in enumerate(allocators): m = d[allocator][perm]["mean"] if not cmax or m > cmax: cmax = m if not cmin or m < cmin: cmin = m maxs[perm] = cmax mins[perm] = cmin fname = ".".join([self.name, "transactions.tex"]) headers = [perm.nthreads for perm in self.iterate_args(args=args)] with open(fname, "w") as f: print("\\begin{tabular}{| l" + " l"*len(headers) + " |}", file=f) print("Fäden / Allokator ", end=" ", file=f) for head in headers: print("& {}".format(head), end=" ", file=f) print("\\\\\n\\hline", file=f) for allocator in allocators: print(allocator, end=" ", file=f) for perm in self.iterate_args(args=args): m = d[allocator][perm]["mean"] s = "& \\textcolor{{{}}}{{{:.3f}}}" if m == maxs[perm]: color = "green" elif m == mins[perm]: color = "red" else: color = "black" print(s.format(color, m), end=" ", file=f) print("\\\\", file=f) print("\\end{tabular}", file=f) self.export_stats_to_csv("transactions") self.export_stats_to_dataref("transactions") mysql = BenchmarkMYSQL()