8 users online. Create an account or sign in to join them.Users
"Archive Overview" Data Source Output Issues
This is an open discussion with 6 replies, filed under Troubleshooting.
Search
This is a major concern for me, and something I have lost much sleep over during the last couple of weeks. I will be working on optimisations over the next week or so specifically for revision 6.
Thank you, Alistair.
By the way: Setting query caching to "on" in config.php like this:
$settings['database']['query_caching'] = 'on';
does not seem to change anything. I am on rev 5. Should this not help?
Hmm, it should force all SELECT queries to have SQL_CACHE added to the start. If you remove the setting from config, it defaults that whatever your server says. If it is set to off, you get SQL_NO_CACHE instead.
If you are not noticing any difference, check your MySQL config. Maybe it isn't enabled on your MySQL server yet? Should have something like this:
query_cache_limit = 1M query_cache_size = 16M
Ah, I see, and I checked this. I am rather sure that MySQL caching is working. Could it be that my problem is not query time, but "XML building time"?
Here are the debug screen's "General Details":
Engine Initialisation 0.0076 s Page creation process started 0.0091 s XML Built 1.0082 s XML Generation 0.0168 s Page Built 1.0282 s XSLT Transformation 0.0498 s Page creation complete 1.0873 s
And here the "Full Page Render Statistics":
Total Database Queries 1246 Slow Queries (> 0.09s) 0 Total Time Spent on Queries 0.0707 s Time Triggering All Events 0.0043 s Time Running All Data Sources 0.981 s XML Generation Function 0.0168 s XSLT Generation 0.0498 s Output Creation Time 1.0873 s
Does the above mean that processing takes so long?
I should mention that I am on a webspace with the following virtualisation limitations:
- 1 CPU core (1.6 GHz) – not allocated dynamically!
- RAM 2.048 MB
- PHP memory limit (global): 32 MB
- PHP memory limit (local): 128 MB
I am glad if you can give me a hint how to optimise this.
It would appear that the Database is coping okay, but the actual generating of those entry objects is taking way too long.
I will try to reproduce your setup on our benchmarking site, add in a few thousands entries, and see where the bottlenecks are. The other concerning thing is the 1200+ queries necessary to produce that XML. At the same time I am looking into how I can optimise the database (was up until 2 am last night pondering/experimenting).
Gimme a couple days. I'll let you know what the best way to go is.
Alistair, just in case it helps your research, I could create a copy of the website on a dedicated development webspace and give you access there. You might change anything you like there.
Create an account or sign in to comment.
I have created a data source which simulates the "Archive Overview" format option available in Symphony 1.7. But I experience a performance issue. Creating the XML for 64 news articles grouped by date takes nearly half a second if I select the entry's title field as output. Here are my values for data source execution:
The data source in question is "newsarchivnavigation" which will output s.th. like:
<news-archiv-navigation> <section id="1" handle="news">News</section> <year value="2008"> <month value="04"> <entry id="731" pdfs="0" bilder="0"> <titel handle="spd-will-bei-dienstrechtsreform-klarheit-noch-vor">SPD will bei Dienstrechtsreform Klarheit noch vor der Wahl</titel> </entry> ... <entry id="650" pdfs="0" bilder="0"> <titel handle="die-csu-schwaecht-mit-ihrer-schulpolitik-die-regio">Die CSU schwächt mit ihrer Schulpolitik die Regionen Bayerns</titel> </entry> </month> <month value="03"> <entry id="642" pdfs="0" bilder="0"> <titel handle="schneider-kapituliert-vor-reform-des-g8">Schneider kapituliert vor Reform des G8</titel> </entry> ... <entry id="423" pdfs="0" bilder="0"> <titel handle="ueberwaeltigender-sieg-fuer-spd-in-nuernberg-und-m">Überwältigender Sieg für SPD in Nürnberg und München</titel> </entry> </month> <month value="02"> <entry id="329" pdfs="0" bilder="4"> <titel handle="maget-stellt-wahre-schuldenuhr-der-csu-vor-fast">Maget stellt wahre Schuldenuhr der CSU vor – fast fünf Milliarden Euro</titel> </entry> ... <entry id="1" pdfs="0" bilder="0"> <titel handle="thomas-beyer-csu-gefaehrdet-hausarztversorgung">Thomas Beyer: CSU gefährdet Hausarztversorgung</titel> </entry> </month> </year> </news-archiv-navigation>If I deselect the field "titel" (which means including no element, thus creating empty "entry" elements), XML generation takes even longer.
I wonder what will hapen with some hundred entries.
Does anybody know a more efficient way to simulate "Archive Overview" data source behaviour?
Are there any plans to re-introduce "Archive Overview" in Symphony 2?