Continuing from the book reading :
This book talks about code efficiency. Performance of code may depend on the machine configuration. For example, a little more memory can help tremendously with page swapping. To speed up a site, there are generally two categories of techniques - namely benchmark and profiling. Benchmarking is experimenting to determine the best approach for something before implementing it for real. Profiling is experimenting on the real thing to see how well it performs. In Benchmarking, there is a timer involved that is started before measuring the duration of some code execution and then stopped right after it. This start-stop is repeated for accumulating the durations. Further the timers could be labeled so that they can be started and stopped independently. The PEAR benchmarking class provides this functionality with its Benchmark_Timer.
Profiling can help you plan the hardware improvements to the existing code. Things like connections, processor, memory can tremendously improve the actual performance.
Web server improvements also will help improve actual performance. Apache is highly configurable. Apache has several httpd processes that run to handle the requests. More processes can be started when the load becomes high and then subsequently the excess can be shutdown. In some cases, these processes may be shut down even if there's work to do to reclaim memory that's been leaking. An httpd process may also crash failing the request but another process can take the same request again. For example, the configuration options to control the number of different httpd processes include MaxClients, MaxRequestsPerChild, StartServers and MinSpareServers. These are specified in the httpd.conf file.
With improvements in hardware, database and Apache, we can focus on code improvements next. PEAR coding standards come useful here.
PEAR provides caching framework Static version of pages can be generated and served instead of regenerating each time. The source code for the page can be stored in pre-parsed format that the Zend engine can readily execute. Even browser cache can be used to improve experience. The LastModified header comes useful to determine the last time the content was changed.
The Expires header can be used to determine how long the page should be valid for. Clients could also use the If-Modified-Since header that can be used by the server to generate a full response or send a Not-Modified response.
Furthermore, output buffering can be used to gather up the contents of the page and send it out all at once. This can be done by including ob_start() and ob_end_flush() through different sections of the html. The ob_start('ob_gzhandler') can be used to make sure the buffered output is compressed.
There are several different caching engines available to choose from. For example, these include Alternative PHP Cache (APC) which is a PHP opcode caching engine and comes with PECL/PEAR, Zend Engine which has also become part of PECL, eAccelerator which is another compiled state PHP caching engine, JPCache which is a memory/disk caching solution and is a PHP library that can be used to store the generated output of a PHP script to disk or in a SQL Server instead of saving the compiled state, and memcached which unlike the output caching systems of APC, eAccelerator and JPCache relies on the caching of backend code objects such as the database result objects and the data-model entities.
This book talks about code efficiency. Performance of code may depend on the machine configuration. For example, a little more memory can help tremendously with page swapping. To speed up a site, there are generally two categories of techniques - namely benchmark and profiling. Benchmarking is experimenting to determine the best approach for something before implementing it for real. Profiling is experimenting on the real thing to see how well it performs. In Benchmarking, there is a timer involved that is started before measuring the duration of some code execution and then stopped right after it. This start-stop is repeated for accumulating the durations. Further the timers could be labeled so that they can be started and stopped independently. The PEAR benchmarking class provides this functionality with its Benchmark_Timer.
Profiling can help you plan the hardware improvements to the existing code. Things like connections, processor, memory can tremendously improve the actual performance.
Web server improvements also will help improve actual performance. Apache is highly configurable. Apache has several httpd processes that run to handle the requests. More processes can be started when the load becomes high and then subsequently the excess can be shutdown. In some cases, these processes may be shut down even if there's work to do to reclaim memory that's been leaking. An httpd process may also crash failing the request but another process can take the same request again. For example, the configuration options to control the number of different httpd processes include MaxClients, MaxRequestsPerChild, StartServers and MinSpareServers. These are specified in the httpd.conf file.
With improvements in hardware, database and Apache, we can focus on code improvements next. PEAR coding standards come useful here.
PEAR provides caching framework Static version of pages can be generated and served instead of regenerating each time. The source code for the page can be stored in pre-parsed format that the Zend engine can readily execute. Even browser cache can be used to improve experience. The LastModified header comes useful to determine the last time the content was changed.
The Expires header can be used to determine how long the page should be valid for. Clients could also use the If-Modified-Since header that can be used by the server to generate a full response or send a Not-Modified response.
Furthermore, output buffering can be used to gather up the contents of the page and send it out all at once. This can be done by including ob_start() and ob_end_flush() through different sections of the html. The ob_start('ob_gzhandler') can be used to make sure the buffered output is compressed.
There are several different caching engines available to choose from. For example, these include Alternative PHP Cache (APC) which is a PHP opcode caching engine and comes with PECL/PEAR, Zend Engine which has also become part of PECL, eAccelerator which is another compiled state PHP caching engine, JPCache which is a memory/disk caching solution and is a PHP library that can be used to store the generated output of a PHP script to disk or in a SQL Server instead of saving the compiled state, and memcached which unlike the output caching systems of APC, eAccelerator and JPCache relies on the caching of backend code objects such as the database result objects and the data-model entities.
No comments:
Post a Comment