Comments on: FreeBSD 6.0 MySQL Performance

By: MySQL i FreeBSD6.0 - dlaczego nie mozna zwiekszyc max_connections | hilpers

MySQL i FreeBSD6.0 - dlaczego nie mozna zwiekszyc max_connections | hilpers — Fri, 23 Jan 2009 04:09:51 +0000

[...] tuningu > kern.ipc.sem*. Znalazlem na sieci ciekawe informacje o wydajnosci MySQL pod FreeBSD. http://blog.gslin.org/archives/2005/12/12/252/ Pomijajac cala dyskusje o wyzszosci Linuxa jako OS dla DB skupilem sie na tuningu jadra. I tak [...]

By: Benchmark for MySQL 5.0 + FreeBSD 6-STABLE

Benchmark for MySQL 5.0 + FreeBSD 6-STABLE — Tue, 26 Dec 2006 07:18:12 +0000

[...] Date: Mon, 12 Dec 2005 12:59:32 +0800 3*2*2*2 = 24 cases: # Compile Options: none, WITH_PROC_SCOPE_PTH=yes, WITH_LINUXTHREADS=yes # /etc/libmap.conf: none (libpthread), libthr # kern.timecounter.choice: ACPI-fast, TSC # kernel: ULE+PREEMPTION, ULE I put the detail information in my blog: http://blog.gslin.org/archives/2005/12/12/252/ — * Gea-Suan Lin (public key: http://ccreader.nctu.edu.tw/~gslin/key.txt) * If you cannot convince them, confuse them. — Harry S Truman _______________________________________________ freebsd-performance@xxxxxxxxxxx mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-performance To unsubscribe, send any mail to "freebsd-performance-unsubscribe@xxxxxxxxxxx" Follow-Ups: [...]

By: Anonymous

Anonymous — Thu, 21 Dec 2006 17:04:22 +0000

[...]

By: Benchmark for MySQL 5.0 + FreeBSD 6-STABLE

Benchmark for MySQL 5.0 + FreeBSD 6-STABLE — Mon, 18 Dec 2006 09:14:30 +0000

[...] Benchmark for MySQL 5.0 + FreeBSD 6-STABLE Benchmark for MySQL 5.0 + FreeBSD 6-STABLE Gea-Suan Lin gslin at gslin.org Sun Dec 11 20:59:35 PST 2005 Previous message: mysql benchmarks Next message: Benchmark for MySQL 5.0 + FreeBSD 6-STABLE Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] 3*2*2*2 = 24 cases: # Compile Options: none, WITH_PROC_SCOPE_PTH=yes, WITH_LINUXTHREADS=yes # /etc/libmap.conf: none (libpthread), libthr # kern.timecounter.choice: ACPI-fast, TSC # kernel: ULE+PREEMPTION, ULE I put the detail information in my blog: http://blog.gslin.org/archives/2005/12/12/252/ — * Gea-Suan Lin (public key: http://ccreader.nctu.edu.tw/~gslin/key.txt) * If you cannot convince them, confuse them. — Harry S Truman Previous message: mysql benchmarks Next message: Benchmark for MySQL 5.0 + FreeBSD 6-STABLE Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the freebsd-performance mailing list [...]

By: bsdforums.org - FreeBSD, OpenBSD, NetBSD, MacOS X, Darwin, Linux, BSD Unix forums, message boards, discussions and news. - mysql stability on fairly busy server

Sun, 17 Dec 2006 22:45:09 +0000

[...] Hello, I’m not taking a deep tought on this, i’m kind of sleeping on my desk (10am here) : It seems to me that you hit a bottleneck, and that’s normal… there is always a final bottleneck, but if you can enlarge this bottleneck so more data can flow in less time, you’ll have better general performance. The question about database tunning is, in my opinion, based in 3 subjects (actually 4 but let’s leave the hardware out of this): 1) Query (application side) tunning and schema design is the highest effect tunning in a database, most people dont believe in this, but you can count on that for sure. Every little improvement in these subjects reflects instantly on response times and resource consuming. 2) Database internal configurations (buffers, caches, pools, indexes, filesystem/raw access, temporary tables, etc). 3) Operating system tunning (filesystem tunning (lower that seek time!), load balance datafiles trough different disks (I/0 spreading), general OS tunnings like SMP and network tunning, threading (Native threading vs LinuxThreads, last time I checked LinuxThreads are still fashion) and other configuration that might help your system performance. There are many, many other optimizations regarding to database tunning, it requires extensive googling, reading, testing and benchmarking. It’s never perfect If you like to read… be sure to take a look at this: http://www.oreilly.com/catalog/hpmysql/ And by the way… i’ve picked up this benchmark some days ago, you can take a look: http://blog.gslin.org/archives/2005/12/12/252/ Cheers __________________ "BSD is, after all, about a commitment to technical excellence, and not about making an operating system platform into a religion" Last edited by nullpt : 06-06-2006 at 04:24 AM. Reason: various typos [...]

By: R_T_F_M

R_T_F_M — Sat, 31 Dec 2005 01:37:43 +0000

FreeBSD 6.0-STABLE #0: Fri Dec 23 21:41:40 CET 2005
rtfm@amd64.home.org:/usr/obj/usr/src/sys/AMD64
ACPI APIC Table:
Timecounter “i8254″ frequency 1193182 Hz quality 0
CPU: AMD Athlon(tm) 64 Processor 3000+ (2010.06-MHz 686-class CPU)
Origin = “AuthenticAMD” Id = 0xf48 Stepping = 8
Features=0x78bfbff
AMD Features=0xe0500800
real memory = 805240832 (767 MB)
avail memory = 778686464 (742 MB)

[my.cnf]
# The MySQL server
[mysqld]
port = 3306
socket = /tmp/mysql.sock
skip-locking
key_buffer = 64M
max_allowed_packet = 64M
table_cache = 512
sort_buffer_size = 2M
read_buffer_size = 2M
read_rnd_buffer_size = 8M
myisam_sort_buffer_size = 64M
thread_cache = 24
query_cache_size = 192M

# Try number of CPU’s*2 for thread_concurrency
thread_concurrency = 8

mysql 5.0.17, libthr, ULE + TSC:
Sat Dec 31 02:30:09 CET 2005
select_index 20000 0 0 25164.77
select_index 20000 0 0 24781.40
select_index 20000 0 0 21857.76
select_index 20000 0 0 21932.90
select_index 20000 1 0 22043.21
Sat Dec 31 02:30:14 CET 2005

mysql 5.0.17, libthr, ULE + ACPI-fast:
Sat Dec 31 02:32:25 CET 2005
select_index 20000 0 0 21618.98
select_index 20000 0 0 22079.83
select_index 20000 0 0 19821.74
select_index 20000 0 0 19771.64
select_index 20000 1 0 19953.37
Sat Dec 31 02:32:30 CET 2005

By: b6s

b6s — Wed, 14 Dec 2005 22:12:23 +0000

average = SUM / 30

998.57 libthr-ule
980.79 libthr-4bsd-pre
915.07 libthr-4bsd
895.54 libpthread-4bsd
889.05 libpthread-4bsd-pre
883.88 libpthread-ule
846.98 libthr-ule-pre
789.92 libpthread-ule-pre

看來我應該做更多 round 再求平均。

By: b6s

b6s — Wed, 14 Dec 2005 21:26:34 +0000

重作所得到的排名 (freebsd 6-stable, mysql41, no pth, tsc)

libthr-4bsd-pre:
select_index 20000 2 0 5165.58
select_index 20000 2 0 5096.83
select_index 20000 1 0 5096.77
select_index 20000 1 0 5116.92
select_index 20000 1 0 5046.69
select_index 10000 4 0 815.11
select_index 10000 4 0 746.69
select_index 10000 4 0 764.19
select_index 10000 2 0 787.74
select_index 10000 3 0 787.14

libthr-4bsd:
select_index 20000 2 0 2507.44
select_index 20000 1 0 5705.03
select_index 20000 1 0 5135.80
select_index 20000 1 0 5179.54
select_index 20000 2 0 5109.68
select_index 10000 3 0 723.18
select_index 10000 5 0 790.10
select_index 10000 5 0 741.24
select_index 10000 2 0 805.16
select_index 10000 2 0 754.84

libthr-ule-pre:
select_index 20000 1 0 2281.18
select_index 20000 2 0 5225.43
select_index 20000 3 0 4679.19
select_index 20000 2 0 4694.60
select_index 20000 2 0 4698.70
select_index 10000 5 0 775.20
select_index 10000 5 0 752.01
select_index 10000 6 1 799.21
select_index 10000 5 0 705.33
select_index 10000 7 1 798.59

libthr-ule:
select_index 20000 2 0 5598.51
select_index 20000 2 0 5089.20
select_index 20000 3 0 5095.78
select_index 20000 1 0 5071.14
select_index 20000 1 0 5124.13
select_index 10000 4 0 819.70
select_index 10000 8 0 769.60
select_index 10000 6 0 782.60
select_index 10000 7 0 825.56
select_index 10000 7 0 780.80

libpthread-4bsd-pre:
select_index 20000 2 0 4843.92
select_index 20000 2 0 4578.51
select_index 20000 2 0 4567.29
select_index 20000 1 0 4563.85
select_index 20000 2 0 4589.73
select_index 10000 7 0 688.09
select_index 10000 6 0 726.53
select_index 10000 10 0 710.14
select_index 10000 7 0 737.06
select_index 10000 4 0 666.36

libpthread-4bsd:
select_index 20000 3 0 4893.14
select_index 20000 4 0 4595.75
select_index 20000 1 0 4604.16
select_index 20000 2 0 4594.44
select_index 20000 1 0 4591.62
select_index 10000 8 0 691.87
select_index 10000 6 0 727.72
select_index 10000 4 0 715.31
select_index 10000 5 0 746.48
select_index 10000 15 0 705.85

libpthread-ule-pre:
select_index 20000 2 0 4351.12
select_index 20000 2 0 4162.48
select_index 20000 3 0 4133.76
select_index 20000 4 0 4128.27
select_index 20000 3 0 3478.31
select_index 10000 6 0 701.57
select_index 10000 10 0 672.38
select_index 10000 7 0 709.34
select_index 10000 11 0 640.43
select_index 10000 10 0 719.84

libpthread-ule:
select_index 20000 1 0 4743.03
select_index 20000 1 0 5211.85
select_index 20000 1 0 4039.21
select_index 20000 2 0 4489.69
select_index 20000 3 0 4466.95
select_index 10000 7 0 714.00
select_index 10000 96 0 734.32
select_index 10000 7 0 708.44
select_index 10000 7 0 693.83
select_index 10000 5 0 715.12

By: b6s

b6s — Tue, 13 Dec 2005 11:32:00 +0000

這是一般系統的預設值（吧）：
mysql41-none-libpthread-tsc-pre-none:
select_index 20000 2 0 2583.29
select_index 20000 1 0 5007.72
select_index 20000 1 0 4981.18
select_index 20000 2 0 4937.11
select_index 20000 1 0 4987.36
select_index 10000 4 0 727.68
select_index 10000 5 0 700.20
select_index 10000 3 0 675.21
select_index 10000 5 0 713.38
select_index 10000 5 0 750.88

不過有兩件事導致我這些實驗可能要砍掉重練：
1. 爽皇想測 libthr 時發現了一件我忘了提的重要事項－－我用了 BUILD_STATIC=yes ……
2. 測試時機器的 loading 不夠穩定，使得 super-smack 在同樣組態下多跑幾次也可能出現極大差異。

By: b6s

b6s — Tue, 13 Dec 2005 08:46:47 +0000

在不開 process scope threading 的情況下，核心使用 PRE+ULE 與否的比較。

mysql41-none-libpthread-tsc-pre-ule:
select_index 20000 2 0 2260.47
select_index 20000 3 0 4619.47
select_index 20000 3 0 4552.68
select_index 20000 3 0 4578.26
select_index 20000 3 0 3851.94
select_index 10000 4 0 714.58
select_index 10000 7 0 775.50
select_index 10000 7 0 772.70
select_index 10000 7 0 707.01
select_index 10000 6 0 709.90

mysql41-none-libpthread-tsc-none-none:
select_index 20000 2 0 4776.43
select_index 20000 2 0 4905.50
select_index 20000 2 0 4881.35
select_index 20000 2 0 4910.65
select_index 20000 1 0 4894.56
select_index 10000 5 0 701.28
select_index 10000 9 0 697.95
select_index 10000 6 0 725.00
select_index 10000 10 0 628.51
select_index 10000 6 0 701.13