Context: It’s a shame to take this one out of context.

Question is “What would make energy and modelling services appealing enough for petroleum companies to invest in?”

Chris Ward (IBM) proposes to make supercomputing available as a service. Fred Busche (IBM) mentions that they’ve got a patent in the works that would cover the application in the oil industry. Adalio Sanchez (IBM, Jam Moderator) says “It just seems like we’re close here and that’s precisely what the Innovation Jam is all about…figuring out how to take the next steps to realize success.” Ward responds that IBM scientists all have access to the technology but

“to get from there to the point where we can sell the business off to Lenovo in maybe 20 years’ time, like IBM has just done with Personal Computers, IBM has to invest in developing the thing. IBM has to know what the clients want to buy. What’s the combination of ‘supercomputer’ (which will do the sums), and ‘know-how’ (skilled IBMers who can make it do the oil field analysis) that’s valuable to the people who have a commercial interest in the answers to the questions?”

If Adalio’s conclusions above are accurate (as a non-IBM’er I have no idea), then one step you might consider is to postpone identifying the immediate barriers, and consider the long term objectives. This is what Chris did to some extent with his response and the comparison to the laptop. Both shared interesting and valuable comments.

From my view, it seems like IBM’s current service strategy is to productize and standardize process driven solutions (ed. note: please refer to this post). This makes good strategic sense for a variety of reasons; the most important being scale and overhead costs. Assuming those constraints don’t radically change in the next 5 years, where would we want the supercomputing program to be then?

If I was a client I’d value having supercomputing services available although the limiting factors now don’t seem to be hardware. Rather, the limitation now is having access to the technology. Sure IBM’ers can have it, but IBM’ers aren’t the ideal test market for the service. External engineers and scientists are, and many can’t gain practical experience needed to learn what the systems are capable of.

IBM has the brainpower, no question, but selling the service to a customer seems like it would require crossing a significant tech transfer bridge. The customer won’t be motivated until they have some fundamentally attractive shared vision for the future outcome.

Unfortunately building shared vision for the future is generally a process in itself. It does not fit conveniently into a product driven cycle, except as an expensive, time intensive, and potentially initiative-killing stage in research and development. It can be made to fit, it’s just not an elegant fit.

One way to work around this issue, gain fast client buy in, and achieve a highly desireable 5 year objective might be to make supercomputing available as a web based service.

In IT analyst speak you might call it a supercomputing ecosystem play. Subsidize the service to make it extremely affordable to those specialists with the capacity and initiative to use it. Screen the applicants, and otherwise give them the run of some share of the system. Fund a support team and an analysis team, to explicitly glean process knowledge from the work of the pioneering programmers.

As early as possible, develop an oil industry vertical, and any others that make short term business sense. Otherwise, let the community of users drive the development. Create conditions in which innovation is enabled to happen elsewhere, and learn from the results. I know, I know… it’s kind of a text book Silicon Valley Blue GeneR (TM) approach. And parts of it might be appropriate and powerful for what you’re talking about.

One of the results is likely to be a best practices hybrid development process that can be productized and standardized for numerous niche markets. I imagine that could be a profitable end state 5 year objective.

[tags]IBM, Innovation Jam, Blue Gene, supercomputing as web service[/tags]