hfss, domain decomposition

Simulating a large finite array, which corresponds to 1 billion unknowns, on a single 2.4GHz Intel Xeon processor PC with 2.0GB RAM in
about 12 hours! WOW! It seems too good to be true. But it has been reported in a paper.

1) What method has been used?
Finite Element with a domain decomposition methodology. The great point is that the method allows for non-conforming/non-matching triangulations across the interfaces. Therefore the mesh generation is relaxed considerably.

2) What does it have to do with HFSS?
Trying to get updated on works done by some of the main HFSS guys,i.e. those who have been active in development of HFSS (inside or outside
ANSOFT) from the beginning, I noticed that they have been focused on the aforementioned area. Particularly I would mention:

2-1) Din Kow Sun: I guess many of you know him for papers he has coauthored with Zoltan Cendes on finite element. Have a look at topics
of his recent research projects in ANSOFT, mentioned at the end of the following paper:

The transfinite-element time-domain method
Din-Kow Sun; Jin-Fa Lee; Cendes, Z.;
Microwave Theory and Techniques, IEEE Transactions on , Volume: 51 , Issue: 10 , Oct. 2003
Pages:2097 - 2105

It reads:
**************
His current research projects include the construction of vector singular bases and applications of domain decomposition and nonconforming finite-element methods (FEMs).
**************

2-2) Jin Fa Lee: This guy is a genius. I believe his great works, when at ANSOFT, resulted in HFSS. Here is the link to his website:

http://esl.eng.ohio-state.edu/indivipage.php?pageID=34

and

http://esl.eng.ohio-state.edu/~csg/

Particularly have a look at the following abstracts:

http://esl.eng.ohio-state.edu/~csg/s...hortcourse.pdf

and the following cool paper (to be published):

A Fast DP-FETI like Domain Decomposition Algorithm for the solution of Large Electromagnetic Problems
By: Marinos N. Vouvakis (from Ansoft), Jin-Fa Lee
http://amath.colorado.edu/faculty/co...s/vouvakis.pdf

It sounds HFSS will have such a feature in future, hopefully, but for near future I would expect a temporary remedy for those who would like
to simulate structures with large number of unknowns on PCs, i.e. addressing more than 2 GB of RAM (most probably 3 GB).

Your points of view would be appreciated.

Regards

finite element method is very powerful tool whne combined with integral method(FE_BI) and DDM.

Can't Microwave Office or ADS replace HFSS? It's cumbersome