Sparx Systems Forum

Enterprise Architect => Automation Interface, Add-Ins and Tools => Topic started by: Daniel Temby on March 19, 2008, 01:23:42 pm

Title: Automation Performance
Post by: Daniel Temby on March 19, 2008, 01:23:42 pm
I am currently using the automation interface from a VBA app to build a large model.  I have found that intially the performance is OK but the longer my import script runs, the slower the performance.

I am currently using SQL Server as my repository.

Are there any tricks to improving performance of automation ?
Do I really need to call update everytime I create an element or connector?

I have also set the following properties on teh repository:

BatchAppend = true
EnableCache = true
EnableUIUpdates = false

But did not see an improvement in performance.

Also I am running version 7.0 Build 818,  are there an performance imrpovements in version 7.1 ?

Thanks

Title: Re: Automation Performance
Post by: «Midnight» on March 19, 2008, 01:32:42 pm
I see some of the same things. In particular I cannot see much effect from using BatchAppend.

Sparxian, others; any ideas?
Title: Re: Automation Performance
Post by: thomas.kilian on March 19, 2008, 08:05:49 pm
I'm using Perl/SQL Server. The performance is (according to my measures) the same even for long runs. However, EA is very chatty with the DB. The build-in cache does not help a bit (it adds up for less than 5% speed improvement). I'm caching elements with Perl (which is quite easy with Perl Hashes) and that gives it a real boost. VB is a bit clumsy to handle here.
Title: Re: Automation Performance
Post by: Daniel Temby on March 20, 2008, 09:46:51 am
I suspect that there is an issue with incorrect or missing DB indexes.

I am running a script that creates 1000's of connectors, perfromance starts at 0.1 second per connector but very quickly degrades. Indicates that the DB is not tuned correctly.

Any ideas ???
Title: Re: Automation Performance
Post by: thomas.kilian on March 20, 2008, 07:17:03 pm
So then. Why not tune it by yourself? Adding indexes just eats disk space (simply speaking). Would be nice to know which additional indexes improved your work.