Academic Facets: News and Views from Spatial's University Community

[#1]

If you do CFD, you need to buy the software, get access to a large computer (the bigger the better) and then hire some IT people to run it (while you do the engineering). So much easier if you could you rent the software you need for a few days, installed and running on a big computer that someone else is maintaining. 

Google,and many others, have shown that in a world of fast, always-on, Internet this approach works surprisingly well (most of the time) for applications like email. Now a company (with links to the University of Southampton, in England) called Deziforce (www.dezineforce.com) is offering online access to analysis packages like Ansys, Fluent and LS-Dyna. Customers don’t buy licenses for software; instead they pay an annual membership fee.

“Softwareas a Service” (SaaS) has been the "next big thing" for sometime, but never happened. Perhaps when the idea was first hyped the Internet didn't have the bandwidth to transfer MB data files, but there were also concerns amongst senior mangers about security.  However interestingly I think the situation is increasingly reversed: in many ways the revolution in connectivity and portable storage has made the cloud a much safer place than the office.

This is apparent on a personal level, when I have files (everything from family photographs to research proposals) that are too important to loose, I deliberately store them in the cloud. If my laptop crashes or the University burns down then it would have little impact (on my computing at least)!  But on a corporate/institutional level it is increasingly obvious that moving data around by email, USB memory stick, or CD, is dangerous (far too easy to leave a laptop on the train). Much safer toprovide online access to the parts of the data, or project, that people need rather than passing copies around. After all online logins have an auditabletrail of access passwords and IP addresses.  

I believe a generation of engineers is now appearing who not only expect to find all their CAD/CAM applications in the cloud, but will probably feel uneasy if they are not.

Deziforce might be in just the right place, at the right time. 

….and how would cloud computing effect geometric modeling kernels? I am not sure but perhaps it will make feasible some approaches to geometric computation thathave always been dismissed as just too slow (i.e. computationally intensive) for practical PC based application.

For example the use of symbolic algebra that provides exact solutions (e.g.. π rather than 3.141592653) has always been dismissed for everyday CAD/CAM applicationas being too slow (although the LEDA and CGal http://www.cgal.org/  libraries use this approach and would argue that there is no problem http://www.mpi-inf.mpg.de/~mehlhorn/ftp/CGALLEDAStMalo.pdf). 

But perhaps if you imagine your 3D modeled running remotely,on a vast parallel computer, these approaches will be viable even for poorengineers using only netbook computers?