Performance MySQL Linux vs. Windows

Hi experts,

we are benchmarking MySQL with PHP client on different operating 
systems. We use a set of small db operations (selects, updates, inserts) 
and measure the number of operations being actioned per second.

The results are astonishing. Linux outperforms Windows on the same 
hardware significantly:

Linux: 3.500 operations/s
Windows: 2.100 operations/s

Spec. details:

Linux: Debian 2.6.24 64bit
Windows: 2008R2 64bit
Hardware: 2x Intel(R) Xeon(R) CPU E5405 @ 2.00GHz, 12 GB RAM, 
Hardware-Raid 1, SATA 750GB
PHP 5.2.10
MySQL 5.1.41/64

Is that a "normal" OS effect or would you assume an installation issue?

Best regards
Keith
0
3/12/2010 11:23:21 AM
comp.databases.mysql 6890 articles. 0 followers. laredotornado (870) is leader. Post Follow

11 Replies
2063 Views

Similar Articles

[PageSpeed] 21
Keith Sauvant wrote:
> Hi experts,
> 
> we are benchmarking MySQL with PHP client on different operating 
> systems. We use a set of small db operations (selects, updates, inserts) 
> and measure the number of operations being actioned per second.
> 
> The results are astonishing. Linux outperforms Windows on the same 
> hardware significantly:
> 
> Linux: 3.500 operations/s
> Windows: 2.100 operations/s
> 
> Spec. details:
> 
> Linux: Debian 2.6.24 64bit
> Windows: 2008R2 64bit
> Hardware: 2x Intel(R) Xeon(R) CPU E5405 @ 2.00GHz, 12 GB RAM, 
> Hardware-Raid 1, SATA 750GB
> PHP 5.2.10
> MySQL 5.1.41/64
> 
> Is that a "normal" OS effect or would you assume an installation issue?
> 
> Best regards
> Keith

Insufficient information.  Have you tuned both systems for maximum 
performance?

-- 
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
jstucklex@attglobal.net
==================
0
Jerry
3/12/2010 11:51:31 AM
Am 12.03.2010 12:51, schrieb Jerry Stuckle:
> Insufficient information.  Have you tuned both systems for maximum
> performance?

Hmmm. Some more information about the MySQL configuration: Parameters as 
recommended by MySQL for servers mostly hosting mysql. All tables MyISAM.

What exactly would you need to know in addition?

Best regards
Keith
0
Keith
3/12/2010 12:03:12 PM
Keith Sauvant wrote:
> Am 12.03.2010 12:51, schrieb Jerry Stuckle:
>> Insufficient information.  Have you tuned both systems for maximum
>> performance?
> 
> Hmmm. Some more information about the MySQL configuration: Parameters as 
> recommended by MySQL for servers mostly hosting mysql. All tables MyISAM.
> 
> What exactly would you need to know in addition?
> 
> Best regards
> Keith

You need to learn how to tune MySQL for best performance for your 
installation.  Every one is different.  And default or recommended 
values are a good place to start.  But they are only the start, not the end.

-- 
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
jstucklex@attglobal.net
==================
0
Jerry
3/12/2010 12:04:36 PM
 > You need to learn how to tune MySQL for best performance for your
 > installation. Every one is different. And default or recommended
 > values are a good place to start. But they are only the start, not
 > the end.

You are absolutely right.

But I am not asking for how to fine tune the last 5 percent out of it. I 
am asking why and if a Windows installation of MySQL normally is >30% 
slower than a Linux installation on the same box.

Anyone?

Best regards
Keith


Am 12.03.2010 13:04, schrieb Jerry Stuckle:
> Keith Sauvant wrote:
>> Am 12.03.2010 12:51, schrieb Jerry Stuckle:
>>> Insufficient information. Have you tuned both systems for maximum
>>> performance?
>>
>> Hmmm. Some more information about the MySQL configuration: Parameters
>> as recommended by MySQL for servers mostly hosting mysql. All tables
>> MyISAM.
>>
>> What exactly would you need to know in addition?
>>
>> Best regards
>> Keith
>

0
Keith
3/12/2010 12:26:44 PM
Keith Sauvant:

>Linux: 3.500 operations/s
>Windows: 2.100 operations/s

>Is that a "normal" OS effect or would you assume an installation
>issue?

Hi Keith,
I have no comparison data, so all I could try to do is point out, that
this question has been asked a lot in several forums already, so you
may want to do a search on "mysql performance linux windows" or similar.
Of course, you'll see plenty of vague opinions and bashing, but I
believe there are a few useful replies out there:

http://forums.mysql.com/read.php?24,215528,228017#msg-228017
http://dev.mysql.com/tech-resources/articles/mysql_on_windows.html


-- 
Erick
0
Erick
3/12/2010 12:43:33 PM
Hi Erick,

thank you for the links. I searched some hours before asking here and 
came across both documents. But as you said, information is all vague 
unfortunately. My hope is to find somebody here with practical 
experience comparing the operating platforms.

Best regards
Keith




Am 12.03.2010 13:43, schrieb Erick T. Barkhuis:
> Keith Sauvant:
>
>> Linux: 3.500 operations/s
>> Windows: 2.100 operations/s
>
>> Is that a "normal" OS effect or would you assume an installation
>> issue?
>
> Hi Keith,
> I have no comparison data, so all I could try to do is point out, that
> this question has been asked a lot in several forums already, so you
> may want to do a search on "mysql performance linux windows" or similar.
> Of course, you'll see plenty of vague opinions and bashing, but I
> believe there are a few useful replies out there:
>
> http://forums.mysql.com/read.php?24,215528,228017#msg-228017
> http://dev.mysql.com/tech-resources/articles/mysql_on_windows.html
0
Keith
3/12/2010 1:19:33 PM
Keith Sauvant <oecher7.z.ksau@spamgourmet.com> wrote:
>
> we are benchmarking MySQL with PHP client on different operating
> systems. We use a set of small db operations (selects, updates, inserts)
> and measure the number of operations being actioned per second.
>
> The results are astonishing. Linux outperforms Windows on the same
> hardware significantly:
>
> Linux: 3.500 operations/s
> Windows: 2.100 operations/s
>
> Is that a "normal" OS effect or would you assume an installation issue?

You benchmark a lot of different things in such a test:

- interprocess communication
- file systems
- scheduler
- compiler + system libs quality

Windows is known for i.e. slow NTFS. And all IPC sucks, that is
why virtually all Windoze software is multithreaded instead of
multiprocessed.

Finally: UNIX (used to be Solaris, nowadays Linux) is the primary
development platform for MySQL developers. No wonder MySQL performs
better there. Oh and don't miss the MySQL manual. It has a Win/*NIX
comparison page too.


XL
0
Axel
3/12/2010 1:49:31 PM
Keith Sauvant spoke:
> thank you for the links. I searched some hours before asking here and 
> came across both documents. But as you said, information is all vague 
> unfortunately. My hope is to find somebody here with practical 
> experience comparing the operating platforms.

I write MySQL clients on this here WinXP laptop, usually connecting to 
our own Linux server for development, and also to client machines and 
hosted databases, through our shonky 2Mb slowband pipe here, and even 
with the crap connection the Linux hosts are noticeably faster than my 
127.0.0.1 connection to the local MySQL on the laptop.



0
R
3/12/2010 2:02:55 PM
Keith Sauvant wrote:
>  > You need to learn how to tune MySQL for best performance for your
>  > installation. Every one is different. And default or recommended
>  > values are a good place to start. But they are only the start, not
>  > the end.
> 
> You are absolutely right.
> 
> But I am not asking for how to fine tune the last 5 percent out of it. I 
> am asking why and if a Windows installation of MySQL normally is >30% 
> slower than a Linux installation on the same box.
> 

Because Microsoft is shit?

Probably disk drivers.
0
The
3/12/2010 3:16:55 PM
R C Nesbit wrote:
> Keith Sauvant spoke:
>> thank you for the links. I searched some hours before asking here and 
>> came across both documents. But as you said, information is all vague 
>> unfortunately. My hope is to find somebody here with practical 
>> experience comparing the operating platforms.
> 
> I write MySQL clients on this here WinXP laptop, usually connecting to 
> our own Linux server for development, and also to client machines and 
> hosted databases, through our shonky 2Mb slowband pipe here, and even 
> with the crap connection the Linux hosts are noticeably faster than my 
> 127.0.0.1 connection to the local MySQL on the laptop.
> 
> 
> 
Ive got the same code running om two different Linux machiones.  The 
faster CPU at the end of a long line is markedly better than the one at 
my knee .

It also has a SATA rather than IDE drive.

I suspect your answer is in the disk subsystem really. Linux is really 
quite good in that area.

0
The
3/12/2010 3:19:40 PM
Keith Sauvant wrote:

> I am asking why and if a Windows installation of MySQL 
> normally is >30% slower than a Linux installation on the same box.

Most likely the ODBC connection on the microsoft install that makes for the
loss of "operations".
-- 

  //Aho
0
J
3/12/2010 4:38:26 PM
Reply:
Similar Artilces:

Strange performance
My OpenGL program runs with almost identical speed no mater what resolution I set (the funniest thing is that it runs slighty faster (+1 FPS) on 1600x1200 than in 640x480). It was tested on GeForce2 and Riva (I don't remember which version). The application uses a variety of things that could cause this (lighting, collision detection, sorting, alpha blending and lots of non opengl stuff). I use depth buffer, no extensions, no draw/read pixel. Where could be the problem? -- Archibald Sounds like you've got your VSYNC enabled in your video driver, which means that no matter h...

Killing sons (Linux)
Maybe this isn't strictly a Ruby question, but I hope someone here can help: I have a job-management application, with a central daemon which receives job requests. Upon receiving this request, it forks and then runs "system" to run bash, which in turn runs the Matlab job. I use bash for this in order to redirect the input and output from Matlab. pstree output looks like this: init-+-apache2---8*[apache2] |-atd ... |-ruby-+-4*[ruby---bash---MATLAB-+-matlab_helper] | | `-15*[{MATLAB}]] | `-{ruby} ... ...

Thunderbird vs Mail
I just switched to OS X. I downloaded Thunderbird because it was based on Mozilla. I like it better than anything I've tried in Classic. How does Mail compare to Thunderbird? Sawney Beane <XXbeadle@qwickconnect.net> wrote: > How does Mail compare to Thunderbird? Seriously, the best comparison is the one you do yourself since only _your_ opinions matter for something like this. -- Mike Rosenberg <http://www.macconsult.com> Macintosh consulting services for NE Florida <http://www.cafepress.com/macconsult,macconsult4> Mac-themed T-shirts <http://bogart-tribute...

[News] Windows Viruses Interfere with Linux-based Device
TomTom shipped viruses on its navigation devices ,----[ Quote ] | The TomTom devices run on Linux, while the two viruses are | Windows-based. Users will be aware that their navigation device | is infected if, for example, they attempt to back up their | content by connecting the device to a PC running antivirus | software. `---- http://news.zdnet.com/2100-1009_22-6154198.html Same old story. Windows has done similar things before (see refs). Related: Small Number of Video iPods Shipped With Windows Virus ,----[ Quote ] | We recently discovered that a small number - less | than 1% - of t...

Oracle Performance Management and Tuning
As a user of Oracle and Linux technology, we are pleased to announce the availability of empower! v2.1 for Oracle. This short note introduces 6 key features that make empower! different from the Oracle desktop tuning tools you're used to. The purpose of empower! is to make all the Oracle tuning facilities and related operations available to you from a single Windows based desktop application. In the past, you might have required a database server login to run essential utilities like STATSPACK and Tkprof - with empower! you no longer need that server login because you can run e...

[News] Review of Audio Applications for GNU/Linux Tribler BitTorrent
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Holiday Cheer, Holiday Uncheer - Part 2 ,----[ Quote ] | Continuing my holiday machine maintenance saga I move on to some notable | trials and tribulations with Ubuntu, but not before I report on a little more | holiday cheer. `---- http://www.linuxjournal.com/content/holiday-cheer-holiday-uncheer-part-2 Tribler: BitTorrent and Beyond ,----[ Quote ] | P2P (peer-to-peer) is the nature of the Net. You can fight that, or you can | embrace it. Here in the US, the mainstream entertainment business has mostly | been fighting it. Hollywood and...

Matching Video Card to CPU performance
I have an AthlonXP 1800+ CPU on a Leadtek 7350KDA m/board - 4x AGP, 266 (DDR) FSB and 512MB DDR SDRAM. I want to replace the current GeForce4 MX440 video card with something faster for gaming. But at what point am I getting too much video card for the CPU/motherboard combination? Would a Radeon 9600XT be wasted on such a combo? Is there a more approporiate 4xAGP card? Any sites that provide references for CPU to Video Card performance matching? Thanks arawata "Andrew" <arawata@ozemail.com.au> wrote in message news:5640b017.0402101836.7957bf96@posting.google.com... > I hav...

JOB: MySQL DBA
Hi, I working closely with an extremely passionate VoIP company in London and we are looking to find a really energetic and lively MySQL expert to fulfil a mix of 2 roles - data architect and database administrator. As they run MySQL RDBMS they would ideally like to find a MySQL expert but they would not say no to someone with Postgres, DB2 or Oracle). Another nice thing about the company is that although they work hard to ensure that they are the most innovative VoIP provider around; they have regular Nintendo Wii battles and you can be sure that although the offices are in the city of Lon...

=?windows-1256?B?ytHPzyDe5MfJIE1CQyDj1dEgx+HMz+3PIDIwMTMtx+E=?= =?windows-1256?B?ytHPzyDH4czP7c8g4d7kx+UgTUJD?=
ytHPzyDe5MfJIE1CQyDj1dEgx+HMz+3PIDIwMTMtx+HK0c/PIMfhzM/tzyDh3uTH5SBNQkMKCmh0 dHA6Ly9uYXRpZ3RhczdhYi5ibG9nc3BvdC5jb20vMjAxMy8wMy9tYmMtMjAxMy1tYmMuaHRtbA== ...

Performance profiling
Hi there, I've written a time library to profile functions using JavaScript without being too intrusive. It works in all the major browsers, IE6+, Firefox, Opera and Safari. You can hook any function and anonymous function using: time.func(myFunction); myFunction(var1, var2, var3); or document.getElementById('link').onclick = time.func(function() { // do something return value; }); Examples of profiling can be seen here: http://remysharp.com/2007/04/20/performance-profiling-javascript/ Hopefully this is of some help - I found it really useful when trying to profile wh...

Best Portal/Home Page for a Linux Advocates? Not Yahoo...
Yahoo has been getting on my nerves lately. For one, they have been adding lots of services (music, news media) that are Winmedia only. Second, their finance section is riddled with articles writen by M$ astroturfers. And today, was the last straw. In their Tech Tuesday section, on Open Source, the first article appearing was about the M$ super computing effort. It mentioned the words "open source" as in "competing with open source". Ok, I get the hint. As a Linux user I'm not wanted in Yahoo. So, I ask -- what should I choose as my new home page? What ...

image sink to bottom of the window?
I have a image buttom to close the popup window. Any easy way to make the image buttom to sink at the bottom of the window? And the buttom should flow with the text/table if the text flow is longer then popup window.... Nick wrote: > I have a image buttom to close the popup window. Any easy way to make > the image buttom to sink at the bottom of the window? And the buttom > should flow with the text/table if the text flow is longer then popup > window.... I found a solution myself. Make a table with height: 100%, and put the image in the last <tr> and make that <t...

sar vs prstat
Hi, I am using sar to monitor the CPU usage on some workstations running Solaris 8. The problem I am seing is that afterwards if I want to see what particular process is taking up the CPU reported by sar (%usr) by using prstat (I have also tried TOP), the usage reported is a lot different. The sum of the CPU used by all processes showed by prstat is a lot lower than the usage reported by sar. Is there some way/tool to correlate the sar results with the CPU usage of each process ? Thanks Pedro pe_ez123 wrote: > Hi, > > I am using sar to monitor the CPU usage on some workstation...

[News] A Look at Puppy Linux and Sabayon Linux 5.1 Review
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Puppy Linux: Just for fun ,----[ Quote ] | Puppy Linux is something different, a tiny | version of Linux that can be stored on a USB | memory drive, will run in memory, and can be | used for working on the move. `---- http://www.itpro.co.uk/618739/puppy-linux-just-for-fun Sabayon Linux 5.1-r1 - Review and Commentary ,----[ Quote ] | It's a good looking desktop and one of the | best and most stable implementations of KDE | 4.x from what we've been able to see so far. `---- http://linuxcritic.com/stories/66-Sabayon-Linux-5.1-r1-Re...

RE: [ace-users] Why ACE memory pool couldn't work well in linux??
This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01C457F9.8AAE972C Content-Type: text/plain Thanks very much for your reply. I have filled and sent PRF to ace-bugs@cs.wustl.edu. In face, bceause ACE_Shared_Memory_Pool is based on System V, so in windows platform I used ACE_MMAP_Memory_Pool and it works well. But in Red hat 7.3, I have tested both ACE_Shared_Memory_Pool and ACE_MMAP_Memory_Pool and they couldn't work both. When I used ACE_MMAP_Memory_Pool, the memory p...

copy constructors hurting performance
I have a String class (I know I am re-inventing the wheel, yes I have heard of boost, and of QString). My copy constructor does a deep (strcpy) of the char *_buffer member. I have a member function func(const String &param). When an actual String is passed as the param this is nice and efficient. I have a constructor which takes a const char* as an argument. And performs a deep copy of the const char *buffer. The issue occurrs when I make the following call: func("this is the param"); The String(const char *buf) constructor is called, making a deep copy of the data. B...

Improve query performance on large table
Hi Folks, I have a very large table containing 170 million rows of coordinats: CREATE TABLE "DB2ADMIN"."COORDINATE" ( "FID" INTEGER NOT NULL , "AID" INTEGER NOT NULL , "X" REAL NOT NULL , "Y" REAL NOT NULL , "Z" REAL NOT NULL ) IN "USERSPACE1" ; ALTER TABLE "DB2ADMIN"."COORDINATE" APPEND ON; CREATE INDEX "DB2ADMIN"."IDX_COORDINATE1" ON "DB2ADMIN&qu...

ATI FireGL V3100 Video card performance issues
I am looking for some assistance with increasing the performance and/or provide GPU priority to ustation, using an ATI FireGL V3100 Video card. The Video card specs are as follows; ATI FireGL V3100 400MHZ 128MB Bios version: 113-A33414-101 Driver version: 8.293.0.0 The PC is an HP workstation XW4200, P4 HT 3.2GHZ, with 1GB RAM, running Windows XP Pro SP2. I notice that when rotating an drawing, with the display mode set to Constant (and graphics Accel turned on), the image rotation is choppy. This is not an overly large drawing. Any recommendations? The RAM usage is abo...

Importing win32api with Windows 95
Sometime in 1998, Mark Hammond wrote: >> import win32api >>ImportError: DLL load failed: One of the library files needed to run >>this application cannot be found. >Your system needs its COM DLLs updated. >You can find the update at the old Pythonwin pages >http://www.python.org/ftp/python/pythonwin/DownloadWin32.html, which also >points to the MS site the update originates from. >Mark. I got the same error today while trying to back-port an application of mine to Windows 95 for a customer. Unfortunately the link pointing to the solution no longer exi...

Canon EOS camera and Linux
I am considering buying a new camera, mostly to take a lot of pictures of various things that I am selling. I am currently looking at Canon EOS and I wanted to know how Linux friendly it is. Is it all based on standard protocols, like USB storage and gphoto2? thanks From: "Ignoramus30024" <ignoramus30024@NOSPAM.30024.invalid> > I am considering buying a new camera, mostly to take a lot of pictures > of various things that I am selling. I am currently looking at Canon > EOS and I wanted to know how Linux friendly it is. Is it all based on > standard protocols, like...

performance
Hi, I have a 4647KB LabView Application running on a laptop (498MHz, 384MB of RAM, 768 MB of Virtual Memory). This application would run and then crashes in the middle and give me a blue screen, and the computer would restart itself. But smaller application about one fourth of the size (1000KB or so) works fine. Does this mean I need more RAM or something? I pull up the task manager and performance is about 30% or less, unless it needs to open a spread sheet, then performance goes to 100%, but then it goes back down to like 5%-30%. Thanks I don't think the application size is a big is...

Solar Winds Orion Network Performance Monitor
Currently, I'm running the 7.2.0 - April, 2004 release of Orion Network Performance Monitor. For the most part, I'm very excited about the application. It's easy to maintain and does a great job of gathering utilization data. However, I am seeing some issues with false ICMP packet loss, as it is displayed on the website, to many devices on my network. It seems to be an application issue. Anyone seen this before? This is a multipart message in MIME format. --=_alternative 007B26EB88256EB4_= Content-Type: text/plain; charset="US-ASCII" That's not FALSE informati...

SVE performance
Now we are using about 100 share varible and still all in office LAN, but takes 3 or 5 mins to initial the share varible link, any setting can improve it? &nbsp; The 'initial' means the PSP LED turning from red to green. They have already been deployed from server side. Have you tried using the Shared Variable Manager like I mentioned in my last post?&nbsp; This would be a good debugging step to take LabVIEW out of the equation. -Justin D Yes, in the share variable manager, it will take 5 mins to see the quality turn from 'connecting, disconnect' to 'good' af...

HADR performance issue
Hi, I have setup a HADR between two servers in different locations. When I compare the application response time in standard environment and in HADR environment, I find the latter is more than twice slower than the former. I have tried to tune some parameters like DB2_HADR_BUF_SIZE, log buffer size, dbheap, but not much improvement. Can anyone give me some suggestions of how to improve HADR performance? We use sync mode of HADR. Thanks. "ebusiness" <wendy_xie@hotmail.com> wrote in message news:1185461168.310732.176250@j4g2000prf.googlegroups.com... > H...

Wanted Linux Sys Admin in San Francisco
We're looking for an experienced sys admin to join our start up in San Francisco. You can learn more about the role and company (including our noteworthy investors) at http://www.wowkast.com. Regards, Tony Lopez Wowkast.com On 2007-10-22, tonylopez <tonylopez@gmail.com> wrote: > We're looking for an experienced sys admin to join our start up in San > Francisco. You can learn more about the role and company (including > our noteworthy investors) at http://xxxxxxxxxxxxxxxxx Will one of your new hire's first tasks be to get your company to stop spammin...