I'm sure that you weren't impugning PHP and I hope I didn't come across
as saying that. My intent in responding was to help contribute to
getting as accurate a set of numbers for you to compare with and also to
note some of the other issues I saw with comparing local performance to
a web server running in the Netherlands, using PHP 4.3. :-)
The points I had were four-fold
1) The 500ms you had noted earlier likely included more TCP latency than
was accounted for and so the 260-fold difference is probably smaller
2) The calculations you had provided were to the last byte, when first
byte, due to output buffering, is probably a more accurate
representation of elapsed execution time on the server, over the
Internet.
3) From my longer-than-should-be-allowed-on-a-mailing-list response,
using a framework has less to do with performance and more to do with
development time and application structure.
4) The longer response times you get from a framework will probably not
impact the end user that much for 95% (arbitrary number) of the web
applications out there. My personal experience with framework-based and
non-framework-based applications has shown me that once you get beyond a
relatively simple application you start to see pretty good dividends in
developer productivity and code maintainability by using a framework.
Kevin
-----Original Message-----
From: web400-bounces@xxxxxxxxxxxx [mailto:web400-bounces@xxxxxxxxxxxx]
On Behalf Of Nathan Andelin
Sent: Friday, September 11, 2009 11:19 AM
To: Web Enabling the AS400 / iSeries
Subject: Re: [WEB400] PHP - Best Appication Structure
From: Kevin Schroeder
A ping isn't a good measurement to use. It represents one packet
whereas a TCP connection requires at least 3 for the handshake.
Thanks for the tip.
I got 343ms from the first SYN to the last ACK before the response
started.
That seems like a better measurement of network latency.
PHP, by default, uses an output buffer so it will not send any packets
to the client until the request has finished executing, or the output
buffer has flushed.
Okay, but the same applies to the ILE interface that I use.
If I measure the time from last ACK to the first response packet
I get 258ms.
If you're using a network sniffer, you might be just getting an ATK
error message, rather than actually testing the application.
I was measuring with Google Chrome, and Fiddler, and seeing actual
results - if that makes a difference.
In order for this to be a proper apples-apples comparison ...
I understand that the comparison is rough. I'm just trying to get a
basic feel for how the ATK framework might be affecting performance -
not to impune php.
-Nathan.
midrange.com
