Eclipse Community Forums: EMF » [EMF/GMF] - Performance issue on load and save model.

Home » Modeling » EMF » [EMF/GMF] - Performance issue on load and save model. (Howto cache model to limit query to the server)

Show: Today's Messages :: Show Polls :: Message Navigator

[EMF/GMF] - Performance issue on load and save model. [message #1063868]

Fri, 14 June 2013 15:21

Philippe EUGENE

Messages: 58
Registered: January 2011

Member

Hello,
I use the : org.eclipse.emf.ecore.uri_mapping to map my logical URI (business URI) to physical URI (HTTP url).
It's a RESTFULL API to save, load and delete EMF/GMF resource. XMI model (EMF and diagram resources) are saved server-side into a database.
Everything works. All EMF mechanisms are completely transparent.

But the number of requests sent to the server is very too important.
With the resolveProxy option à false i have:

A) Loading a GMF diagram => 5 queries
1 query for load the GMF diagram
2 query for load the corresponding EMF model
2 query for load the container of the EMF model

B) Saving a GMF diagram => 9 queries
2 update for saving GMF model and diagram : OK
7 query for reload : GMF resource / EMF model and his container.

As each GMF diagram own it's resourceSet, every openDiagram action do a lot of HTTP and SQL queries.

Then i search a solution to cache model at the client side and then limit query between client and server.

Thanks for help.

Report message to a moderator

Re: [EMF/GMF] - Performance issue on load and save model. [message #1063871 is a reply to message #1063868]

Fri, 14 June 2013 15:39

Ed Merks

Messages: 33140
Registered: July 2009

Senior Member

Philippe,

Comments below.

On 14/06/2013 5:21 PM, Philippe EUGENE wrote:
> Hello,
> I use the : org.eclipse.emf.ecore.uri_mapping to map my logical URI
> (business URI) to physical URI (HTTP url).
> It's a RESTFULL API to save, load and delete EMF/GMF resource. XMI
> model (EMF and diagram resources) are saved server-side into a database.
> Everything works. All EMF mechanisms are completely transparent.
>
> But the number of requests sent to the server is very too important.
> With the resolveProxy option à false i have:
> A) Loading a GMF diagram => 5 queries
> 1 query for load the GMF diagram
> 2 query for load the corresponding EMF model
I wonder why two?
> 2 query for load the container of the EMF model
Perhaps there's a read-only test?
>
> B) Saving a GMF diagram => 9 queries
> 2 update for saving GMF model and diagram : OK
> 7 query for reload : GMF resource / EMF model and his container.
What's being queries?
>
> As each GMF diagram own it's resourceSet, every openDiagram action do
> a lot of HTTP and SQL queries.
>
> Then i search a solution to cache model at the client side and then
> limit query between client and server.
What would you cache? How could you be sure the cache is consistent
with what's one the server?
>
> Thanks for help.
>
>

Ed Merks
Professional Support: https://www.macromodeling.com/

Report message to a moderator

Re: [EMF/GMF] - Performance issue on load and save model. [message #1064617 is a reply to message #1063871]

Thu, 20 June 2013 09:12

Philippe EUGENE

Messages: 58
Registered: January 2011

Member

Ed,
My first improvement was only on the use of options for loading and saving resources EMF.
No need to customize a URIHandler, just (re)-read chapter 15 of the EMF modeling book. Options in PerformanceXMIResourceFactoryImpl (p. 495) works very well.

Now I'll work on the part GMF editor to see if I can reduce the number of queries (especially saving a diagram).
Thanks for your time.
--
Philippe

Report message to a moderator

Previous Topic:	[Teneo] idBag ID generator
Next Topic:	Extending the Ecore meta-metamodel (M3)

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

]

Current Time: Thu Apr 25 06:48:05 GMT 2024

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter