|
GELARDI: Hi, my name is David Gelardi. I'm the Vice President of Deep Computing Capacity on Demand for IBM. And I'm here today to talk about a very interesting project that we have underway with the Exa Corporation.
With me today is Steve Romunday who's the CEO for Exa. Steve, could you talk a little bit about your company before we talk about the project?
ANSWER: Sure. Exa's a software company, we developed the Powerflow software product. And it's a product for simulation of fluid mechanics. And in particular what that means is in for example in the ground transportation industry is simulating the aerodynamics of a vehicle, the thermal management of a vehicle or the acoustics.
So this matters to the vehicle designers in a way for fuel economy, or handling thermal management warranty costs or air acoustics or noise issues and quality aspects of the car. So we're involved in computing this level of physics rather than forcing customers to build physical models and prototypes and testing them experimentally.
GELARDI: So this is another example of moving from physical to digital.
ANSWER: Absolutely. I'll give you an example, one I was personally involved in with one of our major accounts, we were in the design studio up front outside of the US with on the screen a digital image of the car, digital representation of model of the car.
Our results showing the aerodynamically weak spots of the vehicle, and the car is still not meeting its targets for aerodynamics and fuel economy, acceleration and handling. But we're here on a Thursday, and the design review meeting's on Monday to make major decisions about the car, the stylists are just releasing the design on Thursday.
We're in with the stylist on Thursday morning and we're trying to help them figure out what to do. So we're morphing the car live trying to decide what will work aerodynamically and still meet the styling or design intent.
We came up with, on that day, came up with 10 ideas. We then leave the customer, get back to the airport from there, e mail and upload the changes to our facility here in Boston, and call your team and ask for 500 processors worth of equipment to be available Friday afternoon, at Sunday afternoon we ran 10 days, 10 simulations, came back and were able to quantify exactly the impact of each of these design changes.
And Sunday night, from home, we were all presenting these results to the customer early Monday morning their time so that they had quantitative feedback about the design changes to fit the design iteration level.
And that's a major account with major licenses of Powerflow installed already, yet they could not scale up to meet that kind of turnaround time. So traditionally it was, they would go to the meeting with estimates, guesses, ideas, debates of whether these things were mattered or not, and typically then you wouldn't have an impact and potentially the car project would move on not meeting its target.
So we have our major accounts like the automotive OEMs, the large customers, they have in house capacity we handle these peak demands. Second, we have markets like heavy truck or racing, take a NASCAR team or a heavy truck developer, they don't have the same level of volume, and they have equally complex problems but don't have the level of volume to even have these large scale in house servers.
So they've been forced to do without even experimentally without this kind of information for a long period of time. But the challenges are getting so extreme and so difficult that they are forced to change the design process. And we're helping them by giving them access to this supercomputing capacity and only paying for what they need.
If you look at our average demand for the year, that would say we're at 500 processors worth of equipment, but if you look at our peak, it's thousands of processors. So I still need the ability to hit, when customers all come in on Friday with design goals on Monday, I've got to be able to say yes to all of them, not just to one of them, and expect our customers to actually grow in increasing fixed presence and in a variable presence that goes on top of that.
And the fixed presence is there based on, they know they're going to do so many runs and so many simulations per year, and they know they're going to have a certain amount of capacity they're going to use every day.
So they buy that level and it's cheaper to buy to that fixed level capacity than to have it all on demand, and so they lower their costs by doing that and yet still live in an environment that's scalable.
And that I think is, is just unbeatable. When the customers experience this, when they really experience it, they see it on Powerpoint presentations, it's one thing. You know, when they experience it themselves in projects in the [scale up] and being able to deliver when people don't expect any data, is an incredible experience for them, and it's an incredible experience for us to be able to do that.
Why do we partner with IBM, and there's really three attributes from Exa's point of view, and that's capacity. Right? Capacity that we couldn't get access to anywhere else, we couldn't afford to build.
Second is security. And that is paramount for our customers in terms of the data that we're protecting. This is their crown jewels, their most confidential data that we're dealing with, and it has to be kept to the most secure environments.
And third is price, because of the scalability of what is provided, it has to be done in a cost effective way. If we can't do it, if it's not cost effective, customers won't perform. They're not going to do non economic things. So the pricing of this has been key.
| 
