Eclipse Community Forums: Epsilon » [Flock] Large model migration: performance issue

Help

Home

Home » Modeling » Epsilon » [Flock] Large model migration: performance issue

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

[Flock] Large model migration: performance issue [message #1039454]

Fri, 12 April 2013 07:28

ben ben

Messages: 28
Registered: June 2012

Junior Member

Hi everyone,

Migrating my model takes a very long time :
- 15min for 1MB model
- 8h (!!!) for 100MB model

My Computer's specs:
Intel Xeon 3.10GHz Quad Core
8GB RAM

Eclipse.ini:
-Xms256m
-Xmx4096m
--launcher.XXMaxPermSize
512M

Migration is launched using a Flock launch configuration.

That's model loading phase that takes 99% of time.

Is there a way to improve execution time ?

Thanks !

Report message to a moderator

Re: [Flock] Large model migration: performance issue [message #1040303 is a reply to message #1039454]

Sat, 13 April 2013 10:39

Dimitris Kolovos

Messages: 2162
Registered: July 2009
Location: York, UK

Senior Member

Hi,

Is there any chance you could email us a copy of your setup (metamodels, models, transformations) at epsilon.devs@gmail.com so that we can investigate?

Cheers,
Dimitris

Report message to a moderator

Re: [Flock] Large model migration: performance issue [message #1041499 is a reply to message #1040303]

Mon, 15 April 2013 07:16

ben ben

Messages: 28
Registered: June 2012

Junior Member

Unfortunately, I can't provide metamodels because I'm working on a commercial project ...

I have now a bigger problem, Java heap space when migrating.

I let migration work entire week end, I don't know how many hours it has taken (at least 10 ...), but this morning it got that :

Has somebody succeded I migrating large models using Flock ?

Report message to a moderator

Re: [Flock] Large model migration: performance issue [message #1041518 is a reply to message #1041499]

Mon, 15 April 2013 07:41

Dimitris Kolovos

Messages: 2162
Registered: July 2009
Location: York, UK

Senior Member

Hi Ben,

We have used Epsilon languages with rather large models and although it is slower than Java, 15 mins for a 1MB model is nowhere near what I'd expect. Could you perhaps uninstall Concordance (About->Installation Details->Model Monitoring->Uninstall) and give it another go in case it somehow interferes with the model loading process? If you keep observing the same behaviour could you perhaps try to reproduce the problem with an artificial metamodel and models you can share? Another - clearly more heavyweight - option might be to consider sharing your metamodels/models over an NDA (we've done plenty of these in the past).

Cheers,
Dimitris

Report message to a moderator

Re: [Flock] Large model migration: performance issue [message #1041811 is a reply to message #1041518]

Mon, 15 April 2013 16:01

Louis Rose

Messages: 440
Registered: July 2009
Location: York, United Kingdom

Senior Member

Hi Ben,

Thanks for reporting this issue. I'm going to be working on Flock over the next few weeks, and I'd be really keen to address the performance issues that you're experiencing. If it's possible for you to share some artificial large models with us for which Flock's performance is poor, that'd be really helpful. Alternatively, we could go down the NDA route, as Dimitris suggests.

Our email address is epsilon.devs@gmail.com.

Cheers,
Louis.

Report message to a moderator

Re: [Flock] Large model migration: performance issue [message #1041822 is a reply to message #1039454]

Mon, 15 April 2013 16:12

ben ben

Messages: 28
Registered: June 2012

Junior Member

Hi Louis,

I'm trying to find someone in our dev team that can past some hours "obfuscating" our metamodels.

I probably will be able to give you that before the end of this week.

Report message to a moderator