Maybe you’ve enabled the “flash slideshow” within the Slideshow options of NextGEN gallery. This will load the JW Image Rotator on every page. I’ve disable usage of the flash slideshow by unselecting the option:

[ ] Integrate the flash based slideshow for all flash supported devices

Without usage of the flash image rotator it is still possible to use slideshows. But the overall page load time increases. For me it looks like that loading the flash object causes the overall page display to be delayed. Google Chrome seems to handle loading of the flash object faster.

Anyway for me the fastest browser available seems to be Google Chrome. On my 13″ MacBook I recognize significant performance differences when using Safari, Firefox and Chrome. That’s the reason why I’m currently using Chrome (development version) as browser.

Delete all revision of every post. It can be achieved using a simple SQL statement:

DELETE a,b,c
FROM wp_posts a
LEFT JOIN wp_term_relationships b ON (a.ID = b.object_id)
LEFT JOIN wp_postmeta c ON (a.ID = c.post_id)
WHERE a.post_type = 'revision'

I do this from time to time. Optionally it’s possible either to disable revison managment or restrict the number of saved revisions by adding one of the following statements into your wp-config.php:

define('WP_POST_REVISIONS', false );

define('WP_POST_REVISIONS', 3);

The information has been taken from weblogtoolscollection.com.

This is nice, but WordPress does still send a few HTTP requests to netdna.yoast.com. I think it makes no sense because we have disabled the Yoast widget:

To avoid this few unneeded request I’ve installed the Yoast Remove plugin. After installing and enabling the plugin the additional HTTP requests are gone and as a result we’ve improved the load time our admin dashboard a little bit:

Links

• Yoast Remove plugin

A simple and free tool for private usage is available from simplie Software. It lets you run short and long benchmarks of your hard discs and compare the result with other hard discs from a database.

For me it is a sufficient tool as I wanted to compare the overall performance of two different hard discs.

There is also a online shop to buy the software for commercial use.

Links

• HD Tach Homepage

• Improve Windows Vista Performance

Basically the article describes 4 things which help to improve the performance:

1. Part 2: Improve boot time using Windows ReadyBoost
2. Part 3: Disable Windows Aero
3. Part 4: Use extended hard disc write cache
4. Part 5: Disable unused hardware

On my Windows 7 installation I followed item 1. and 3. Think I have to make further checks how much it improves performance.

Especially item 1. ReadyBoost sounds interesting: Windows uses a USB stick or flash memory card to as additional cache. But I don’t want to use a USB stick, I want to use a PCI Express flash card adapter with a very fast memory card. Guess this should bring best performance.

After first checks my Windows 7 installation boots in about 1:45 minutes (completely). Looks not so bad, since I’ve already installed software like Photoshop, Office, iTunes, etc. The real boot time of Windows 7 is slightly lower than BIOS start up also needs time, approx. 7 seconds.

A very useful site for photographers comes from Rob Galbraith. A lot of CF/SD cards are measured with many different camera models regarding performance. Also several card readers are measured.

Also interesting for speed freak: the Pro Dual CompactFlash Adapter ExpressCard. The adapter allows to put 2 CF cards inside and offers very high performance.

If you want to have it really fast have a look!

Links

• CF/SD Performance Database
• Pro Dual CompactFlash Adapter ExpressCard

Basics

But there are a few things which could help to improve your performance. On my configuration I’ve added skip-name-resolve in my.cnf:

[mysqld]
skip-name-resolve

Now the MySQL server will not do any DNS name resolves. Please note that permissions based on hostnames will no longer work!

For security reasons I would also recommend binding only to the localhost address. Using this the MySQL server is not reachable on the Ethernet interface. In earlier version of MySQL this was usually done with the skip-networking parameter.

[mysqld]
bind-address    = 127.0.0.1

And I’ve added skip-external-locking and skip-locking:

[mysqld]
skip-external-locking
skip-locking

This could also improve your performance.

Search Length

For text search queries with a allowed length of 2 characters add this line:

ft_min_word_len="2"

Otherwise default minimum length is 3 characters, which means that a search query “select * like ‘%AB%’” will never return any data.

Cache Settings

Caching seems to be a very important thing when optimizing your server. My feeling is that the standard configuration of a MySQL server is not really designed for machines with 1 GByte memory or above.

Below is my configuration, with a few cache settings increased.

[mysqld]
query_cache_limit    =  1M
query_cache_size     = 32M
thread_cache_size    =  8
join_buffer          =  1M
table_cache          = 16M
sort_buffer_size     =  2M
read_buffer_size     =  2M
read_rnd_buffer_size =  2M
tmp_table_size       = 48M
max_allowed_packet   = 16M
thread_stack         = 128K
key_buffer_size      = 32M

This configuration works on a machine with 1 GByte RAM and several databases running. Basically I recommend to increase the cache size settings. And memory helps almost always.

Applies to MySQL version: 5.0.32

apt-get install iperf

Optionally the source or the Windows binary is available from sourceforge.net/projects/iperf

Building and installing from source is done using a few commands:

tar xvfz iperf-2.0.4.tar.gz
cd iperf-2.0.4
./configure
make
make install

Once installed iperf the server is started on machine 1:

machine1:~# iperf -s
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------

Next the client is started on machine 2:

machine2:~# iperf -c machine1.mydomain.de
------------------------------------------------------------
Client connecting to machine1.mydomain.de, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[  3] local xx.xx.xx.xx port 45325 connected with yy.yy.yy.yyport 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  96.6 MBytes  80.9 Mbits/sec

Tested on two servers connected with 100 MBit/s located at two different providers in Germany we get a result of 80.9 MBits/sec.

This result is slightly higher compared to the result from our measurement done with NETIO in the article Network Throughput Measurement. So we start the test again with addtional parameters:

machine2:~# iperf -c machine1.mydomain.de -w 512k -l 512k
------------------------------------------------------------
Client connecting to machine1.mydomain.de, TCP port 5001
TCP window size:   256 KByte (WARNING: requested   512 KByte)
------------------------------------------------------------
[  3] local xx.xx.xx.xx port 40987 connected with yy.yy.yy.yyport 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  72.0 MBytes  60.3 Mbits/sec

Now using modified read/write buffer size and TCP windows size we get almost the same result as in our test with netio.

Definitely a tool which can be recommended to check your network throughput!

The example below shows a sample measurement between to servers connected with a 100 MBit ethernet card and located at two different providers.

On one machine NETIO is started as server:

machine1:~/netio/bin# ./linux-i386 -s
NETIO - Network Throughput Benchmark, Version 1.26
(C) 1997-2005 Kai Uwe Rommel
TCP server listening.
UDP server listening.
TCP connection established ...
Receiving from client, packet size  1k ...
Sending to client, packet size  1k ...
Receiving from client, packet size  2k ...
Sending to client, packet size  2k ...
Receiving from client, packet size  4k ...
Sending to client, packet size  4k ...
Receiving from client, packet size  8k ...
Sending to client, packet size  8k ...
Receiving from client, packet size 16k ...
Sending to client, packet size 16k ...
Receiving from client, packet size 32k ...
Sending to client, packet size 32k ...
Done.
TCP server listening.

On the other machine NETIO is called e.g. with -t option for TCP and the name or IP of the other machine:

machine2:~/netio/bin# ./linux-i386 -t machine1.mydomain.de
NETIO - Network Throughput Benchmark, Version 1.26
(C) 1997-2005 Kai Uwe Rommel
TCP connection established.
Packet size  1k bytes:  7309 KByte/s Tx,  7645 KByte/s Rx.
Packet size  2k bytes:  7276 KByte/s Tx,  7724 KByte/s Rx.
Packet size  4k bytes:  7291 KByte/s Tx,  7737 KByte/s Rx.
Packet size  8k bytes:  7276 KByte/s Tx,  7723 KByte/s Rx.
Packet size 16k bytes:  7286 KByte/s Tx,  7719 KByte/s Rx.
Packet size 32k bytes:  7291 KByte/s Tx,  7725 KByte/s Rx.
Done.

We take a rough average of the result can multiply it by 8: 7,5 * 8 = 60 MBit/s. This is a acceptable value for a connection between two servers located at different providers within Germany.

For sure a useful tool! Also check Network Bandwidth Performance Measurement with Iperf