[frascati] Re: Dynamical creation components and wires

[Lionel Seinturier <Lionel.Seinturier@xxxxxxxxxxxxxx>]

Hi Sergey,

On 08/10/2010 18:06, kerzolster@xxxxxxxxx wrote:
> My question can seem a little silly, because I have recently 
> started to use SCA and FraSCAti implementation.
>
> Anyway, Is it a possible to create a lot of (~10 000) components and a lot of
> wires between them (~100 000)?

[1] reports on benchmarks which have been done with FraSCAti. Up to 
13,000 components have been created with FraSCAti in a JVM which was 
allocated with 64MB of memory. The total number of wires was in the same 
range than the number of components.

> And how I can create and destroy them at the
> runtime? 

Before answering this question, let me ask another one, as the answer 
may differ: what is your starting point for creating these components? 
Raw component classes? Or .composite XML files in the SCA Assembly Language?

> Another words, I want to use components like actors
> (http://en.wikipedia.org/wiki/Actor_model).

Great.

> And probably the last question: 
> Somebody did tests with a a large number of
> of components and communications? 
> What about performance, scalability

As far as I know, no, but in the end this depends a lot on the 
communication stack that you use to communicate, from intra-JVM local 
method calls, to remote calls with let say, Apache CXF SOAP Web Services.

> and safety?

Do you mean safety in presence of high throughput communication, or 
safety in general? Anyway, in both cases, I'm not aware of such 
experiments with FraSCAti, although the team of Frantisek Plasil at 
Charles University in Prague, used to have done work on behavioral 
protocols with Fractal components. Given the proximity of Fractal and 
FraSCAti, the results should be easily applicable to FraSCAti too.

Cheers,
Lionel.

[1] L. Seinturier, P. Merle, D. Fournier, N. Dolet, V. Schiavoni, J.-B. 
Stefani. Reconfigurable SCA Applications with the FraSCAti Platform. 6th 
IEEE International Conference on Service Computing (SCC’09). Pages 
268-275. Bangalore, India. September 2009.

--
Lionel Seinturier
Univ. Lille - Lab. LIFL - INRIA ADAM
http://www.lifl.fr/~seinturi