[Preface / Disclaimer: I haven't yet had the pleasure of working closely with Cadence's own Hosted Design Solutions team, so the following will likely reveal ignorance of strategies and solutions that they already have in place to address the issues outlined below. However, given the ideas this panel has inspired, you can be sure I'll be closing this personal information gap in the very near future ...]
A surprise, last minute addition to the DVCon 2009 program was a panel on Software As A Service ("SaaS") as it applies to EDA. (The last 4 frames of the set pointed to by this link are photos of this event.) Organized and moderated by Harry "The ASIC Guy" Gries, the panelists were all SaaS veterans who shared their experiences in a series of opening remarks, and who were very patient in fielding Q&A from the largely novice audience. Over the course of the event, the following issues, some of which are unique to EDA, bubbled to the surface:
* Classes of applications that clearly benefit from SaaS
* Bandwidth needs
* Configuration control
* Dealing with and/or migrating legacy flows & data
Before commenting on each of these in turn, allow me to cut to the chase with the following unscientific, gut feelings that:
* None of the above issues will block significant adoption of SaaS by the EDA industry, within the panelists' general consensus time frame of ~5 years.
* EDA applications that easily leverage multiple CPUs (metric driven verification, anyone?) will surely be in the vanguard of this movement, but this is not to say that all other EDA tools won't eventually migrate to the cloud too.
* In many ways, it's not terribly important to me what platform my products run on, as long as someone wants to run them! ;-) Seriously, whether it's a local farm of Linux boxes, Suns, AIX machines, or some remote, virtualized cloud of processing power, it's the applications themselves (and methodologies that help users extract the most value from said applications) that's most important to the end user.
In future posts, I'll step back and address each of the SaaS concerns listed above in the context of EDA in general, and verification in specific.