• 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
shopping for a scalable php framework

#11
[eluser]edoornav[/eluser]
Thanks for all of the responses. I tried XCache and eAccelerator and got up to 75 rps (from 20). This is probably fast enough that it won't be the bottleneck in a large application.

Has anyone used this framework to build an application that had to scale to more than one application server or db server? What were your findings? Were there any parts of the framework that you couldn't afford to use? Thanks.

~rvr

#12
[eluser]bosco500[/eluser]
[quote author="woopsicle" date="1192216440"]Hi gunter,

I am actually the creator of styleignite. There is no CI caching at all. The speeds you see are a results of eAccellerator and some optimized mysql query cache settings... and a pretty good VPS server.

There's always room for improvement ofcourse so if traffic were to be solidly high then I would probably need to look into view caching of some sort - although it is a bit difficult on a purely dynamic site.[/quote]

I stumbled across Styleignite today, that site rocks! Nice job

#13
[eluser]woopsicle[/eluser]
cheers - traffic has died off a bit now. delicious was great but for some reason noone wanted to digg it.

just need to see how to entice people to upload styles.. it is proving harder than I thought!

very happy how CI held up with the first day flood of traffic.

edoornav: sorry no experience with scaling CI across multiple servers - i can't imagine any reason why it wouldn't hold up though.

#14
[eluser]Michael Wales[/eluser]
I saw it featured somewhere, smashingmagazine I think.

#15
[eluser]kucerar[/eluser]
[quote author="edoornav" date="1192188666"]
static hello world page -> 6500 rps
<?php print "hello world" ?> -> 4000 rps
retrieve one db row -> 1750 rps

CI hello world -> 170 rps
CI retrieve one db row -> 106 rps

[/quote]

Content caching solutions aside, this is kind of shocking. I would expect these kinds of numbers from Symphony, not CI. Can you turn on the benchmarking class to find out what's taking all the time?

Thx
-R

#16
[eluser]kucerar[/eluser]
[quote author="kucerar" date="1192258729"][quote author="edoornav" date="1192188666"]
static hello world page -> 6500 rps
<?php print "hello world" ?> -> 4000 rps
retrieve one db row -> 1750 rps

CI hello world -> 170 rps
CI retrieve one db row -> 106 rps

[/quote]

Content caching solutions aside, this is kind of shocking. I would expect these kinds of numbers from Symphony, not CI. Can you turn on the benchmarking class to find out what's taking all the time?
[/quote]

This guy got 600 rps for CI after using eAccelerator on his hello_world():

http://www.alrond.com/en/2007/jan/25/per...rameworks/
http://www.alrond.com/en/2007/feb/04/in-...rameworks/

Wonder what it would take to write a faster MVC, the best are all 10 times slower than static page.

#17
[eluser]nmweb[/eluser]
Using a framework always cuts down on performance. Ci's approach to load only the stuff you need minimizes this. Only thing lacking is the caching of parts of a page, this could boost performance in a number of cases.

Anyway, keep in mind that what's most important is the loading speed for the end-user. And 80% of this is taken account for after the loading of the html, so after CI is done with it. See for more information this page

#18
[eluser]gunter[/eluser]
by the way, has someone tried Amazon EC2? you know, where you can set up a new server within 3 minutes on demand

#19
[eluser]edoornav[/eluser]
[quote author="kucerar" date="1192258729"][quote author="edoornav" date="1192188666"]
static hello world page -> 6500 rps
<?php print "hello world" ?> -> 4000 rps
retrieve one db row -> 1750 rps

CI hello world -> 170 rps
CI retrieve one db row -> 106 rps

[/quote]

Content caching solutions aside, this is kind of shocking. I would expect these kinds of numbers from Symphony, not CI. Can you turn on the benchmarking class to find out what's taking all the time?

Thx
-R[/quote]

I turned on the profiler in the constructor of my controller. It says:

Loading Time Base Classes: 0.0064
Controller Execution Time ( Blog / Index ): 0.0087

This seems quite fast. I turned on top and noticed that during this benchmarking, apache processes where using all of the CPU. APC, etc seem to help a lot here. On my laptop I went from 20 rps to 100 rps. A similar increase on our dev server (I don't want to install APC on it right now) would be over 500 rps, which is really good.

~ryan

#20
[eluser]kucerar[/eluser]
[quote author="edoornav" date="1192444869"]

I turned on the profiler in the constructor of my controller. It says:

Loading Time Base Classes: 0.0064
Controller Execution Time ( Blog / Index ): 0.0087

This seems quite fast. I turned on top and noticed that during this benchmarking, apache processes where using all of the CPU. APC, etc seem to help a lot here. On my laptop I went from 20 rps to 100 rps. A similar increase on our dev server (I don't want to install APC on it right now) would be over 500 rps, which is really good.

~ryan[/quote]

Thanks for your report...Maybe the problem is not so much the controller implementation as your numbers indicate, but just the include processing overhead etc.

Will have to check out APC. Right now stuck with Zend core, would have to take my chances plugging in APC or eAccelerator, or upgrading to the enterprise performance platform to get the bytecode caching.

-R


Digg   Delicious   Reddit   Facebook   Twitter   StumbleUpon  


  Theme © 2014 iAndrew  
Powered By MyBB, © 2002-2020 MyBB Group.