Friday 22 June 2012 9:33:10 am - 3 replies
I have a potential customer asking me very detailed question about eZ publish (they have a tech staff that will be doing work on eZ), and one of the issues coming up is documentation. Specifically about performance benchmarking.
Can anyone give me some pointers on where to find documented information on this issue?
Friday 22 June 2012 11:06:16 am
For benchmarking page-load times, all the usual tools apply, from Apache Bench to Funkload, passing by Jmeter. Nothing really specific.
For benchmarking the innards of page execution, the "ezperformancelogger" extension on projects.ez.no is your friend: it allows to trace different perf metrics for every page access. Both the trace format and the metrics logged are configurable - they include total time, ram, number of queries and time spent on them etc...
For benchmarking the rest of the system, usual suspects apply as well: free, vmstat, iostats, htop, mytop, mtop, munin, collectd etc...
For keeping an eye on perfs while developing, enable the debug ouput and be sure to check it, even when you have to scroll at the bottom of the screen to see it.
The ezdebug_template_operator, also on projects.ez.no, allows you to easily add timing points and counters from template code, so that you can benchmark specific pieces of code.
To do the same from php code, look at ezdebug::accumulatorstart
Last but not least: come to the eZ summer Camp, where I will be holding a workshop on the topic!
Modified on Friday 22 June 2012 11:07:34 am by Gaetano Giunta
Monday 25 June 2012 9:19:33 am
I probably have to rephrase a bit
Do eZ provide any official docs on performance/benchmarking?
How about comparisons to other similar systems? Urls etc.
This is basically docs my potensial customer wants as a piece in the puzzle to base his platform decision on.
Modified on Monday 25 June 2012 9:22:33 am by Kristian Blom
Tuesday 26 June 2012 8:00:05 pm
I do not think that eZ provides any official docs on performance/benchmarking - but you should ask your sales representative for a definitive answer.
The main problem is that the "framwork" nature of the CMS makes it hard to compare benchmark results - it is easy both to get stellar numbers and horrible ones...
What we normally do in our sales materials is use numbers from reference customers - and we have many with really huge PVs ...
Modified on Tuesday 26 June 2012 8:01:12 pm by Gaetano Giunta
You must be logged in to post messages in this topic!