<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML dir=ltr xmlns:o = "urn:schemas-microsoft-com:office:office"><HEAD><TITLE>Message</TITLE>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<META content="MSHTML 6.00.2900.2802" name=GENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=#ffffff>
<DIV><FONT face=Arial size=2>Hi,</FONT></DIV>
<DIV> </DIV>
<DIV><FONT face=Arial size=2><FONT face="Times New Roman" size=3>We performed
some model scaling experiments recently, which might be of<BR>interest to some
of you. These point to potential limitations of GME with<BR>respect to the size
of models that can be scaled at once in batch. We were<BR>using the latest
publicly available version of GME for this.<BR><BR>There is a Petri Net modeling
language that Dr. Andy Gokhale created (called<BR>the SRNML). In this language,
an "event type" for the Reactor pattern is<BR>modeled as a series of places and
transitions. We wanted to use our model<BR>transformation engine (called C-SAW)
to push the GME to its limits in terms<BR>of the number of modeling elements
that could be added.<BR><BR>We created model transformations to scale from 2
event types to 438 event<BR>types. This represents about 4,400 modeling elements
(basically, all of<BR>these were atoms and connections) added in batch all at
once. On the<BR>addition of the 439 event type, the GME crashes.<BR><BR>We
confirmed this situation on several different machines, so it seems to be<BR>a
characteristic of GME that is independent of the platform.<BR><BR>An interesting
observation was made: it was possible to add 500 event types<BR>if the
replication was performed as two steps of 250 event types (rather<BR>than 500 at
once, which would crash at 438). This seems to suggest that a<BR>GME interpreter
may add modeling elements at a speed that eventually GME<BR>could not process.
It would be good to get feedback on this from GME<BR>developers on this
observation.<BR><BR>Part of this was done for a journal paper we just submitted
on model<BR>transformations as replicators. In case anyone is interested in
looking at<BR>this work, an URL for the submitted paper (currently under review
by SoSym)<BR>is at:<BR><BR></FONT><A href=""><FONT face="Times New Roman"
size=3>http://www.cis.uab.edu/gray/Pubs/sosym-models-2006.pdf</FONT></A><BR><BR><FONT
face="Times New Roman" size=3>Thanks<BR><BR>Jane, Jeff, and Jing at
UAB</FONT><BR></DIV></FONT>
<DIV>---------------------------------------------------------------------------------<BR>Yuehua
(Jane) Lin<BR>Department of Computer and Information Sciences<BR>University of
Alabama at Birmingham<BR><A href="mailto:liny@cis.uab.edu">liny@cis.uab.edu</A>
<BR><A href="http://www.cis.uab.edu/liny">http://www.cis.uab.edu/liny</A></DIV>
<DIV> </DIV>
<DIV><FONT face=Arial size=2></FONT> </DIV></BODY></HTML>