DDDAS - Dynamic Data Driven Application Systems

From the desk of Samy,

What is DDDAS? It is a new paradigm, where computer programs act in a more intelligent way. Their outputs depend on their inputs, which are updated values in the real world, used for optimizing the minimum damage/etc of important problems such as natural catastrophes. The other day, Dra. Frederica Darema came to UAB to give us a conference about DDDAS. I attended it. We could say that DDDAS is a new direction for applications/simulations and measurement methodology.

Until now, programs are mostly unidirectional. It means that we have a direction between theory and simulation, and ANOTHER ONE, between simulation and theory. In the same way, we’ve got a third element: measurements.

This is too static and serialized. Too slow! This is enough for normal programs and situations, but what happens when LIFES are up to a few minutes? We have to speed it up as much as we can. So, this isDDDAS: Data is not unidirectional but bidirectional. Or in other words, we haven’t got a strong direction, we’ve got a data stream going and coming. A stream data between execution processes.

An example of how effective and important this can be, we could prevent tornados better and faster, and know where they will go.

So, the current paradigm is a program which receives parameters/measurements constantly, but with no communication between source and destination.

DDDAS is different. DDDAS receives inputs that have been asked previously… by destination before. Let’s create the scenario to understand it better.

From the desk of Samy,

What is DDDAS? It is a new paradigm, where computer programs act in a more intelligent way. Their outputs depend on their inputs, which are updated values in the real world, used for optimizing the minimum damage/etc of important problems such as natural catastrophes. The other day, Dra. Frederica Darema came to UAB to give us a conference about DDDAS. I attended it. We could say thatDDDAS is a new direction for applications/simulations and measurement methodology.

This is too static and serialized. Too slow! This is enough for normal programs and situations, but what happens when LIFES are up to a few minutes? We have to speed it up as much as we can. So, this isDDDAS: Data is not unidirectional but bidirectional. Or in other words, we haven’t got a strong direction, we’ve got a data stream going and coming. A stream data between execution processes.

An example of how effective and important this can be, we could prevent tornados better and faster, and know where they will go.

So, the current paradigm is a program which receives parameters/measurements constantly, but with no communication between source and destination.

DDDAS is different. DDDAS receives inputs that have been asked previously… by destination before. Let’s create the scenario to understand it better.

OLD PARADIGM: We’ve got a machine (source) that measures wind constants, among others constants. A program (destination) receives those dataand processes it. An imaginary time to prevent a natural catastrophe: 40 minutes.

NEW PARADIGM: A measurement machine gets wind constants, etc. Destination receives it. But, oh, there’s something particular in that data that will probably give some problems (are we gonna have a tornado soon? or a tsunami?) depending on which are the other constants. Unfortunately, our measurement machine is not programmed for sending this information all the time because it’s very expensive. But this new paradigm will solve our problems, because destination will ask this information and the source will get that, and go ahead forwarding it to the source. So, this is the start of a loop. Destination will ask the information that it really needs all the time, because other data will probably be dummie now. So, inputs depend on the outputs of the destination. Inputs depend on the outputs, and it’s asynchronous. A imaginary time to prevent a natural catastrophe: 8 minutes.

The challenges of DDDAS fall on the design and the architect to do so, because this is fantastic but requires a lot of power.

Nowadays, there are a lot of researches in this area. One of them, a department of IBM.

In 1980, Dra. Frederica Darema came at these ideas, but computers weren’t ready for that.

“DDDAS has the potential to revolutionize science, engineering and management systems” – Darema 2000.

I’ve choosed the tornado example, but you can imagine the applications in these fields:

1. Engineering

2. Medical purposes

3. Manufacturing

4. Business

5. …

DDDAS is yet a beta version. We don’t need to solve 100% of the problems, but kind of.

A scheme of problems:

• Application development (layer)

• Algorithms (tolerant to perturbations of dynamic input data)

• Measurements

• Systems supporting such dynamic environment

Now we’ve got the resources to create a good DDDAS, now it’s time. Computers have the enough tech to run this type of things. We’ve got also a model which merge multiple heterogeneous systems(GRID).

So, the main difference between DDDAS and the old paradigm is the capability to incorporate data to an executing process. In other words, DDDAS means dynamic data integration!

Another example is the fire model. Now, information is based on statistics and we need time to process it. But using DDDAS, we no longer need statistics because information is in real time.

Another example: when a military pilot has to eject, for torsion forces or whatever, we don’t know this at the very moment, we know this after 10 seconds. With DDDAS, 3 seconds probably. Or less.

Till now, we had instruments and sensors on the one hand, and computer platforms (GRID) on the other hand. With DDDAS, this is all the same, because there is a stream among all processes.

DDDAS is the future world. DDDAS is for a better world.

…SaMy*^33