In a recent blog post titled A revolution in EDA?, I talked about how small EDA companies might be able to use modern web technologies and trends to help them reduce the high cost of sales so pervasive in EDA right now. In that post, I mentioned that EDA users are looking for better solutions, with good ROI, and flexible licensing terms. I talked about a number of things related to this, but one thing I didn't cover was the licensing and distribution model. People have commented that "Software as a Service" (or SaaS) might be a good model for the future. Of course, there's a lot of different notions of just what constitutes SaaS, but one commonly referred to example is Salesforce.com, where the customers' data and the application software is hosted offsite from the customer premises and the customer accesses the software on demand via a web browser.
Now there are a variety of advantages to SaaS implementations like Salesforce.com. Here is an abridged list, taken directly from Salesforce.com's website describing what they believe are the top benefits of on-demand software:
- Unparalleled time to value
- Less expensive initially
- Easy upgrades
- Better service delivery
- Better scalability
- Easier to customize
- Users are more satisfied and productive
- Easier for administrator
- Nurturing true value and innovation
Now I'm sure that other vendors of non-SaaS software would argue about some of these points and further might even bring up their own list of advantages, but there's no denying that SaaS carries with it a certain set of advantages.
From my perspective, the primary customer-oriented advantages of SaaS are:
- Less expensive initially due to low upfront licensing costs
- Reduced risk of adoption due to incremental usage and billing and low upfront cost
- Ease of administration, encapsulating ease-of-installation and ease-of-upgrade.
- Better scalability, due to the fact that the customer can increase or decrease software usage without the purchase of new machines or new software licenses and without the headaches of administration of a larger number of machines and licenses
- Ease of cross-geographic and/or mobile access to the application and data
The above list seems pretty compelling, but I think EDA distribution and licensing model based on a Salesforce.com like model is a long way off. I have several reasons for this belief:
- Performance. Most serious EDA applications require high performance CPU and IO performance. Think about place&route, layout, functional verification, analog simulation, etc. Pushing multi-gigabyte data sets through the internet just doesn't work very well. Local installation, where performance and compute resources are under user control, is a major plus in this area.
- Interoperability with other tools. When a piece of software is hosted, it becomes much more difficult to interoperate with other tools through scripted flows. Scripting is a fact of life in every design I’ve ever worked on. So again, local installation becomes a major factor.
- Data security. From my experience, companies are very gun-shy about letting any design data off of their corporate network. For most big EDA users it is simply out of the question.
For most users of performance-heavy EDA applications, I would be willing to bet that a monetary accounting of the value of SaaS does not compensate for the disadvantages that I pointed out. But I think there may be room for some sort of hybrid approach that allows for some of the advantages of SaaS while mitigating some of the disadvantages. One possibility is to have a per-user monthly billing. Another possibility is metered usage, similar in concept to how cell-phone contracts work. In each of these cases, the upfront cost is reduced with a concomitant reduction in the perception of risk. There are also many things that can be done in the software implementation to reduce the administrative overhead of installation and maintenance.
All that being said, there may be some subset of customers who are interested in a SaaS EDA offering. On top of that, there is an undeniable trend towards SaaS and cloud computing. So perhaps the best advice for EDA startups is to architect your software and business model so that you are not caught flat-footed if SaaS becomes a requirement from your customers.