Download Speed

[Drew***]

While our forum is happily running on phpbb2. We have a second version on phpbb3 so that we can develop our style, learn our way around, and play.

I began to hear from a few members that had visited the new version that it was very nice, but a bit slower. Since I never noticed that phpBB3 was slow, and even thought phpBB3 was a bit faster, I started asking questions. Processing speed was not the issue since we are on our own dedicated server and nowhere near capacity. It seems that everyone commenting on the slower times has dial-up or a limited (< 256K per sec ) speed connection. Running this http://www.websiteoptimization.com/serv ... com/phpBB/ I think that that the CSS file may be the reason. Is there any way around this? We will enable zlib.output_compression and see if that helps; any thoughts on this?

URL
1 67689 CSS style.php?ampid=2&amplang=en
1 24939 HTML http://area51.phpbb.com/phpBB/
1 4334 SCRIPT styles/prosilver/template/forum_fn.js
1 3430 IMG styles/prosilver/imageset/site_logo.gif
1 3046 CSS* styles/prosilver/theme/print.css
1 2517 SCRIPT styles/prosilver/template/styleswitcher.js
7 186 IMG styles/prosilver/imageset/icon_topic_latest.gif
1 26 CSS styles/prosilver/theme/normal.css
1 26 CSS* styles/prosilver/theme/medium.css
1 26 CSS* styles/prosilver/theme/large.css
7 ^ 103121* Total (^unique objects)

edit * meant zlib.output_compression - sorry, too much going on

[SamG]

I understand from a couple of similar discussions that there is some sort of CSS bottleneck and that RC4 will address the problem.

[Grok]

I had similar complaints and after considerable head-scratching found that in my case the problem was users that were using IE7 with the Prosilver style. Switching them over to Subsilver2 eliminated the complaints.

[Drew***]

I ran a comparison and the differences were impressive

Download Times*
Connection Rate Download Time
14.4K 95.42 seconds
28.8K 48.41 seconds
33.6K 41.69 seconds
56K 25.58 seconds
ISDN 128K 8.80 seconds
T1 1.44Mbps 2.04 seconds

*Note that these download times are based on the full connection rate for ISDN and T1 connections. Modem connections (56Kbps or less) are corrected by a packet loss factor of 0.7. All download times include delays due to round-trip latency with an average of 0.2 seconds per object. With 7 total objects for this page, that computes to a total lag time due to latency of 1.4 seconds. Note also that this download time calculation does not take into account delays due to XHTML parsing and rendering.
Page Objects
QTY SIZE# TYPE URL
1 67696 CSS http://www.********.com/phpBB3/style.php?sid=1328b7e975adc02cc928f22320de1aee&ampid=1&amplang=en
1 43056 HTML http://www.********.com/phpBB3/index.php
1 4334 SCRIPT http://www.********.com/phpBB3/styles/prosilver/template/forum_fn.js
1 3490 IMG http://www.********.com/phpBB3/styles/prosilver/imageset/site_logo.gif
1 3046 CSS* http://www.********.com/phpBB3/styles/prosilver/theme/print.css
1 2517 SCRIPT http://www.********.com/phpBB3/styles/prosilver/template/styleswitcher.js
32 186 IMG http://www.********.com/phpBB3/styles/prosilver/imageset/icon_topic_latest.gif
1 26 CSS http://www.********.com/phpBB3/styles/prosilver/theme/normal.css
1 26 CSS* http://www.********.com/phpBB3/styles/prosilver/theme/medium.css
1 26 CSS* http://www.********.com/phpBB3/styles/prosilver/theme/large.css
7 ^ 121305* Total (^unique objects)



The results after changing php.ini settings.



Download Times*
Connection Rate Download Time
14.4K 24.48 seconds
28.8K 13.04 seconds
33.6K 11.41 seconds
56K 7.48 seconds
ISDN 128K 3.40 seconds
T1 1.44Mbps 1.76 seconds

*Note that these download times are based on the full connection rate for ISDN and T1 connections. Modem connections (56Kbps or less) are corrected by a packet loss factor of 0.7. All download times include delays due to round-trip latency with an average of 0.2 seconds per object. With 8 total objects for this page, that computes to a total lag time due to latency of 1.6 seconds. Note also that this download time calculation does not take into account delays due to XHTML parsing and rendering.
Page Objects
QTY SIZE# TYPE URL
1 12965 CSS http://www.********.com/phpBB3/style.php?sid=e01304623dafe2f60faf3c5f496a43c5&ampid=1&amplang=en
1 5965 HTML http://www.********.com/phpBB3/index.php
1 4334 SCRIPT http://www.********.com/phpBB3/styles/prosilver/template/forum_fn.js
1 3490 IMG http://www.********.com/phpBB3/styles/prosilver/imageset/site_logo.gif
1 3046 CSS* http://www.********.com/phpBB3/styles/prosilver/theme/print.css
1 2517 SCRIPT http://www.********.com/phpBB3/styles/prosilver/template/styleswitcher.js
32 186 IMG http://www.********.com/phpBB3/styles/prosilver/imageset/icon_topic_latest.gif
1 43 IMG http://www.********.com/phpBB3/cron.php?cron_type=tidy_sessions&ampsid=e01304623dafe2f60faf3c5f496a43c5
1 26 CSS http://www.********.com/phpBB3/styles/prosilver/theme/normal.css
1 26 CSS* http://www.********.com/phpBB3/styles/prosilver/theme/medium.css
1 26 CSS* http://www.********.com/phpBB3/styles/prosilver/theme/large.css
8 ^ 29526* Total (^unique objects)

# Congratulations. This site is using HTTP compression, otherwise called content encoding using gzip. The sizes reported here are for compressed content sent from the server to the client.

* CSS alternate stylesheets may be referenced in the HTML but are not actually downloaded until they are needed and are therefore not included in the total page size.

These are only benchmarks and results will vary in the real world, but this resolves the issue for us. However this may not do much to help those who have no control over their php.ini settings

[removals]

I still don't understand:(

[Drew***]

Forgive me if this is a stupid idea, these days I only keep up with things enough to make them work, and sometimes when the fish are biting I don't even do that.

Noticing that "CSS alternate style-sheets may be referenced in the HTML but are not actually downloaded until they are needed " got me to wondering if the CSS could be divided so that only those elements needed would be called up. One solution could be to use different headers say forum, upc & mcp but I'm sure there are other more elegant solutions.

[Prince of area51]

hmm one way I handled this issue with one of the websites I did was to actually tell the template engine which css files to place in the <head> tag.

[Highway of Life]

There are a couple of possible solutions here that you could implement.

1. A new lighter slower-connection style. -- all users with dial-up should use this style.
   You could go to a very basic style using few to no images and very little CSS. -- though this requires a new writup of CSS -- though not much need to touch the templates if done right.
   I’m thinking something similar to an Archive look that you see in many BBs today. -- not pretty, but it is certainly fast.
   If you do this, I would set all your Bots to this theme specifically... they will like it much better.
2. Separate the CP specific CSS into their own files, and take those sections out of stylesheet.css
   Add

   Code: Select all

   <!-- IF S_IN_MCP --><link href="{T_THEME_PATH}/mcp.css" rel="stylesheet" type="text/css" media="screen, projection" /><!-- ELSEIF S_IN_UCP --><link href="{T_THEME_PATH}/ucp.css" rel="stylesheet" type="text/css" media="screen, projection" /><!-- ENDIF --> 

   Though you have to remember to keep the images in the main stylesheets since those CP specific stylesheets are not going to parse the image variables.

[Drew***]

Thank you for the suggestion

Several of the boards I admin are aimed at the interests of retirees. As you might guess most won't spring for a high-speed connection (at least until they discover that they can flood the internet with photos of their granchildren) and modifying templates mainly consists of stripping features that cause confusion. For example it is easier to strip "The team • Delete all board cookies • All times are UTC" from the breadcrumbs template than to keep trying to explain what these things mean. This is by no means a complaint. I'm very happy that the forum is adaptable.

One especially important Kudo for whom ever is responsible, prosilver is much easier to read for many older adults especially with the ability to change font size.

[code reader]

what would be interesting to see, imo, is some comparison between loading speeds of prosilver vs. subsilver2.
if, as i suspect, the problem is prosilver specific, it may give some clue to a possible solution.

since you ran all those measurements, i wonder how much bother it would be to run them again, this time with subsilver2?
thx.