Reaching Mom Tactic: Analogies

“Sweater, n.:  garment worn by child when its mother is feeling chilly.” – Ambrose Bierce

The whole point of an analogy is to explain something new by comparing it to something known.  The trouble with analogies is they are inherently imperfect and can lead to incorrectly assuming properties of the known also apply to the new.  Additionally, the known may be so comfortable to the audience (your Mom) that it superimposes itself on what you do.

Cutting Computing Architectures Down to Size

As part of my job, I regularly talk about advances in computer technology – most notably processors and computers built with multitudes of processors. But why is this hard? Can’t you just buy “the best” processor from Intel or IBM or NVIDIA? The challenge is the myriad of different problems we try to solve with a computing platform – mapped to a complex universe of possible solutions across combinations of hardware and software choices.

After many flawed analogies, the simplest I’ve come to employ is yard work. Many of our Moms assigned yard work chores while we were growing up. On any given Saturday, two tasks I may have performed were mowing the lawn and cutting down an old tree. Abstractly, these are the same task: employ a machine to sever plant material.

I was willing to spend more time, care, and fuel on the tree task than a single blade of grass. Thus, a chainsaw is the tool of choice for the job.  But to cut the lawn with a chainsaw would result in poor quality, wasted time and fuel, and perhaps cause the neighbors to hide behind shuttered windows! Similarly, while a mower is time-efficient at cutting grass (a very large number of blades of grass cut simultaneously in a “massively parallel” process), it lacks the capability to chop down a tree.

So what do Computer Architecture and yard work have in common? You need to understand the variety of tools, particularly as new ones are invented, and then properly select and apply them to the required task. The risk is overhearing Mom then repeat: “He is a computer gardener” – but the payoff is eerily lucent: “He is trying to invent a lawnmower that works on forests.  But it’s really computers and data.”

Putting Cloud Computing Through the Wringer

The buzz around “Cloud Computing” is so pervasive, even Mom asks what’s the big deal. My favorite analogy first appeared in Christofer Hoff’s Rational Survivability blog: Laundry. You have a home computer for data. You have a home washer /dryer for clothes. They are designed to carry a workload proportional to expected historic use at a point of need. There is a clear value in knowing your washer is available and that the intimates you put in it stay in the house. But you had to purchase the washer and dryer, make an informed choice in doing so, and you expect it to work for many years.  Most likely you do not employ all of its features and over that time do not benefit from advances in washer/dryer technology improvements.

Now suppose you host a family reunion and suddenly the demand for clothes washing spikes, either you can inefficiently employ your domestic appliances or load up baskets and either drive to a Laundromat (Infrastructure-as-a-Service/IaaS) or have these picked up by a Laundry Service (Software-as-a-Service/SaaS). The advantages here include: someone with more expertise than yourself selected the appliances, purchased them, maintains them, and you only pay when you use them – for the small part of their life you use. Because their purpose is to serve a market of users (multitenancy) there is a much larger capacity collectively (and perhaps even individually) than at home.

So rather than running 8 loads one after another, you can stuff 4 larger machines at once and complete the task in 1/8^th of the time. If a laundry service, you even benefit from their expertise in operating the machines and using detergents, and offloading the labor involved in the process from dirty to wash to dry to fold.

However – there are some inconveniences and risks. You need to be able to pay at time of service (perhaps with a bucket of quarters), you incur a delay in the movement of your clothes to and from these machines, others are using the machines, so it’s possible you may need to wait or that your intimates may be seen by others if care is not taken, or you may even lose something in the process.

There are many flaws to this model – it is incomplete, exaggerates some aspects, and clothes are not digital (yet) so cannot be replicated or transmitted (like a virtual closet). But as a canonical task often lovingly delegated to Mom (particularly in the college years), laundry is a familiar experience from which to discuss “The Cloud.”  This analogy is also ironic and potentially confusing on two fronts: One, GE obviously manufactures actual washers/dryers, and two, GE Aviation builds computers for aircraft that literally operate in the clouds.

Pleasing Mom

There is a clear benefit to this exercise, no matter how tedious or seemingly futile. We, as passionate practitioners of engineering and science, directly benefit from being able to clearly communicate our work to non-technical people. We need the ability to describe how our work is important to our employer and customers, what we actually do, and why it is challenging. The fact that our Mom wants to get the low down on our highly technical job, and has the patience to listen is actually a gift and great practice.

I lead a Computing lab that frequently collaborates with Mechanical Engineers, Physicists, Biologists, Chemists, etc., so it’s not unusual that a courageously asked naïve question actually leads to a novel approach at problem-solving.  Our discussion with Mom becomes an “outside” viewpoint that forces us to think about a technical problem from a radically different angle.  These outside viewpoints can lead to insights and connect us with new colleagues (which should please Mom as she always wants you to make new friends.)

At Your Mother’s Knee

One of the greatest gifts of Mothers is a strong foundation from which we build everything we become. While our technical skills may not have come from Mom, we can thank her for fostering us being curious, observant, disciplined, and patient. To then reach into the darkness where nobody has before imagined, taking the calculated risks needed to reap great reward – we are ever-armed with the confidence, the safe harbor, and the encouragement of our Mothers.  Happy Mother’s Day to all the Moms out there and I hope your day is spent on a cloud that needs no explanation — cloud nine.

 “No one in the world can take the place of your mother.
Right or wrong, from her viewpoint you are always right.
She may scold you for little things, but never for the big ones.”  – Harry Truman

This past November, I attended the Supercomputing 2012 Conference in Salt Lake City. Each year, this conference attracts around 10,000 people from around the world and is the venue where many companies make big announcements about new products or capabilities.  The conference is jointly sponsored by the Association for Computing Machinery (ACM) and the Institute of Electrical and Electronics Engineers (IEEE), the two most prominent computer-related professional societies in the world.

This year, the Conference began with a keynote speech from Dr. Michio Kaku, CUNY physics professor and PBS/Discovery Channel futurist. Past keynote speakers for this conference included Al Gore, Bill Gates, Michael Dell, and futurist Ray Kurzweil.

While Dr. Kaku discussed a range of topics from $0.01 computers (greeting cards that play music have as much computing power as the Allied Forces in WW2) to smart toilets, I was most taken by his appeal to the audience to step up to promote science and advocate popular and political support for scientific pursuits.

For example, the financial and human impact of the most recent “Frankenstorm” to hit the Northeast US would have been far more devastating without the advanced warnings and preparations made possible by data collected from satellites and weather models to predict the path and magnitude of the effect.  Yet when budgets are allocated to science and engineering projects like climate analysis, supercomputers, space programs, etc. we as scientists and engineers have shied away from delivering the message on the very practical and clearly economically beneficial results of the same past investments. NASA seems to be learning however, and now has a very active Twitter presence (@NASA), even having the Curiosity Rover wish a Happy New Year to Times Square.

Further, he painted a future where knowledge workers would be critical – that computing will fall into the backdrop of our lives just as electricity or water supply is now. To take advantage of such a world, we will need a much more computer-savvy workforce, and we need to begin promoting more STEM (science, technology, engineering and math) programs, particularly in computer science & engineering, earlier into our children’s education. He suggested we all get more engaged with our school boards, speaking out to our friends and relatives, and advocating to our elected representatives the importance and impact of science & technology investment on our prosperity.

A few additional events at SC12 have shifted this topic’s importance in my mind, particularly conversations with Debra Goldfarb (pictured with me below in the Enterprise captain’s chair.) On my panel on the Future of Manufacturing, we discussed the nature of manufacturing jobs in a world of 3D printers and additive manufacturing.

An Intel panel that included vocational and community college participants discussed workforce development and the needed pipeline of talent and student interest in pursuing careers leveraging computing, modeling and simulation as manufacturing shifts to computer-guided automation.  To expand the skilled workforce with computer programming aptitude, we need to reach students earlier in their studies to plant the seeds of that interest.

Intel invited Dr. Lazaro Lopez, Principal of Wheeling High School located northwest of Chicago to tell the story of the success of a public school integrating STEM into its curriculum with dramatic positive results on student success, confidence, morale, and in particular placement into jobs and colleges despite comparatively humble family incomes in the district. This school integrated into its community, partnering with local firms to offer hands-on work with tangible impacts – increasing the feeling of relevance of topics that often seem purely academic.

As a reader of Edison’s Desk, you must have a warm spot in your heart for Science and Engineering, as Project Lead The Way reaches more schools like Wheeling, and reflecting on the future needs for the workforce as computers blend into so many aspects of the world around us, perhaps we should all consider how we can be advocates of technology and in particular computer science and engineering in our schools, as a career to our young relatives, and the importance of programs like NASA to our common future.

In the upcoming weeks, I will be attending the Workforce Development Institute conference, focused around training high school graduates with the skills needed for future jobs, with clear emphasis on the prominence of STEM across traditional fields. My hope is such tools will help our kids and future generations live long and prosper.

Upcoming Blog Topics:
High Performance Computing: What does Exascale mean?
Cloud Computing: How to explain The Cloud to your Mom
And more Science-as-Art – stay tuned!

The purpose of this event is to bring together world experts and to educate the GE Global Research staff in important technical areas.  The symposium encourages the free exchange of ideas between the speakers and the GE technical staff. Some topics from previous years include Bionics, Energy, Health Care, Sustainability, Networks, Services, Manufacturing, and Imaging & Therapy.

This year the Whitney Symposium discussed Analytics, Big Data, Modeling and Simulation, High Performance Computing and the Industrial Internet as key technological enablers of GE’s Software initiatives. We enjoyed speakers from prestigious universities like Berkeley, MIT, Harvard, and CMU joining government experts from Livermore Lab and DARPA as well as industrial giants like IBM and Procter & Gamble for this year’s discussion.

Bringing together GE researchers and leadership to meet with thought-leaders from academia, government and industry helps us position ourselves at the forefront of these disruptive technologies. The breadth of GE’s products and services lends plenty of opportunity to apply new ideas, learn from them and then magnify them across other industries. Software is a common underlying element to nearly everything we do, and these topics: advanced analytics and modeling, simulation of systems, application of advanced computing technology and the growing interconnectedness of people, software and devices are key to turning imagination into reality.

Below are a few of my takeaways from the symposium. Any thoughts around these? For others who attended, what were your key takeaways?

- We are experiencing a shift from practicing empirical physical sciences to using such studies merely to validate results of ever-advancing high-fidelity simulations.

- Exciting new advances in machine learning and artificial intelligence will augment human intuition and perception in interpreting the “Big Data” being generated through the myriad of scientific and industrial internet applications that have been emerging over the past years.

- There are  possibilities for new discoveries in biomedical and materials science by leveraging collaborative data sharing and advanced modeling and simulation to focus experimentation in the practice of those disciplines.

- While hardware advances emerge and accelerate in computing and network technologies, software skills are more important than ever to truly harness the power of the Industrial Internet and the computational infrastructure behind it.

Several of the Whitney Symposium speakers pose for a picture at Edison’s desk.

I had the chance to talk with Brian to provide you with a bit more insight on the below image, beyond its visual appeal. At the bottom of the post, I’ve also shared an animation, so be sure to check that out as well. See my Q&A with Brian below and feel free to ask questions in the comment box regarding the visual and the work we are doing in CFD.

This image is hypnotizing. What are we seeing and how does it help GE?
What we are visualizing is the unsteady flow through the low pressure turbine (LPT) of a jet aircraft engine.  The flow field is fundamentally unsteady due to the unsteady geometry, i.e. the blades are alternatingly rotating or stationary, but historically our computational technology was limited to steady flow. By computing the unsteadiness, we get a better understanding of aerodynamics loss which we can utilized to design more efficient products.

We used an in-house code called TACOMA to run this simulation – tell us about how it came to be and what advantages we have thanks to our solver?
GE has been utilizing CFD for the development of our turbomachinery products for 30 years.  As computer power has increased, the complexity of the problems we want to simulate increases and this requires continuous attention and development of the CFD solver. In order to push the state-of-the-art in the space of turbomachinery, we have found value in having close control of our software, and so we decided many years ago to write our own CFD solver. TACOMA is our latest code and it has been in use for over a decade.

What other numerical tools are used in the design of turbomachinery?
A code like TACOMA is part of an ecosystem of software that address the need to create geometry, meshes, and to visualize the results. While much attention gets paid to the results of unsteady 3D analysis, we also rely on 1D and 2D tools as well. These lower order tools are very useful during the preliminary design phase when geometry is still being laid out.

How do we make sure a computer model gives us a good result compared to a physical test?
Validation is an important topic that cannot be ignored when utilizing simulations. GE has a variety of physical data from rig tests and full engine tests. We routinely compare our simulations back to our data to make sure we understand their accuracy.

What advantages are there to virtual engineering test over physical tests?
Simulations and physical tests are both important and both have a role in the design process. Simulations are much faster to run (compared to building test hardware) and give more data than testing. This makes simulations well suited for optimizing a design and for digging into flow physics. Physical tests are important to validate the design and to ensure we understand the accuracy of the simulations.

The Discovery Channel shared a short video featuring John Hengeveld, the marketing director for High Performance Computing (HPC) at Intel discussing how computation can play a role in curing diseases. I recommend checking it out to better understand HPC’s role in science, engineering and medicine!

The next image in the third part of our supercomputing “Science as Art” series is an image we titled, “Chromatic Ring”

This image is a modern design of a Low Pressure Turbine and a visualization of the computational geometry that served as input to a Computational Fluid Dynamics simulation.

The unsteady Low Pressure Turbine calculation uses a grid that is hundreds of times bigger than the grids used for a typical design calculation and  computer processing time thousands of times greater.  Thus it requires a supercomputer.  This visualization was run at Oak Ridge National Laboratory on their supercomputer named Jaguar.  Earlier this year, Global Research acquired our own Cray supercomputer, which allows us to now run such experiments on site on our own systems.  We are interested in such calculations because we are trying to find subtle effects that would allow us to improve the Low Pressure Turbine efficiency.

I’m back again with the next image in our “Science as Art” supercomputer-generated series.  The following two images we call, “ExcaLESbur” and “Lava Lamp” (respectively.)

The first image, “ExcaLESbur” features density gradient contours from a Large Eddy Simulation (LES) solution of exhaust flow from a dual flow conic nozzle.  LES is a numerical approach where the turbulence in the flow is solved for directly instead of being modeled as in traditional methods.

The second image “Lava Lamp” shows vorticity contours from a LES  solution of exhaust flow from a dual flow conic nozzle.  Vorticity is a way of quantifying the amount of “circulation” or “rotation” in a fluid element.

These simulations were run at Argonne National Lab on BlueGene/P system. It was run on 8192 Cores with 150M grid points. This work is trying to improve designs of jet engines to reduce noise while minimally impacting other desirable attributes like fuel efficiency. The animation of this image shows a time-accurate flow prediction of jet engine plume exhaust.  Density gradient contours are shown to highlight both the turbulence eddies in the jet shear layer and the acoustic waves propagating from the jet.  Another way to characterize the flow features is by looking at the vorticity contours. As the jet plume develops one notices that the region near the nozzle exit is dominated by small energetic vortical structures (blue). These vortical structures become bigger as the flow moves downstream.

The simulation attempts to capture the noise generated by the exhaust plume of a jet engine.  As anyone who has ever heard a plane fly overhead can confirm, aircraft engines generate a significant amount of noise, and jet noise is a dominant component of the noise.  This is created by high velocity jets and the turbulent structures generated in shearing flows.  The noise is broadband in nature and extends over a large range of frequencies.  While traditional approaches model the turbulence in the flow field, the LES approach solves for the desired range of turbulent scales directly.

These photos show how the LES approach directly captures a large range of turbulent structures, and also the acoustics emanating from these structures.  The traditional approaches have been limited to canonical geometries, a simulation methodology that captures the turbulence phenomena accurately can create a design capability that has the potential to deliver advanced “green” energy and propulsion systems.

Stay tuned.  The next image up in the series is called “Chromatic Ring!”

I wanted to introduce today’s post and a theme that is not new to the blog, but will be getting some additional attention going forward.  That theme is “Science as Art.”  Many times our work in the scientific and engineering communities generates images that are visually breath-taking.  Whether these are computer generated or captured through advanced instruments, whether turbulent flows, cell visualizations, or molecules at the Nano-scale, the fact that images from the scientific community can be beautiful beyond revealing information  is undeniable.

Over the next few months Edison’s Desk will feature other “Science as Art” examples.  Today we are sharing some images that were generated by supercomputers.  A few interesting things to note about these photos: First off, we listed the number of grid points in each image.  A grid point is a vertex in the geometry of the model.  For example, a triangle would have 3 “grid points” and a cube would have 8 “grid points.”  The more grid points, the more complex the shape.

We’ve also listed the number of cores used to compute the data from which each image was generated.  For those of you without a background in computers, I wanted to explain what this meant to give it a little more context.  Each core is one processor in the computer.  For example, many laptops have a “dual core” (2 cores) processor.  Getting programs to work on a very large number of cores at the same time is extremely difficult, so having a high number of cores, say 8 or 64 or 512, is something of note.  The top programs in the national labs run on hundreds of thousands of cores!  Impressive!

And now on to the photos…

The first image I am going to share is lovingly titled, “Rice Krispies® Treats”

This image is a computer simulation of gamma-prime precipitate microstructures in Ni-base Superalloys.  From left to right: (1) Spherical precipitates formed at a low lattice misfit between precipitates and matrix (2) Cuboidal precipitates at an increased lattice misfit -0.3% (typical in single crystal superalloys), and (3) rafting microstructure developed under uniaxial tensile loading.

Run on an 8 core x86 system at GE Global Research in Niskayuna. 16M grid points.  The goal of this work is to understand properties of an alloy based on its molecular structure. The images are from a set of phase field simulations by varying the material parameter (lattice misfit) and mechanical loading condition. They show the effect of these parameters on precipitate microstructure, which eventually control the material’s mechanical properties. The rafting microstructure (the rightmost image) is often observed in single crystal Ni-base superalloys used in high pressure turbine blades (airfoils or buckets) under service conditions. The particular lamellar rafted structure is considered beneficial to increase creep life due to their blockage to dislocation movement under mechanical loading.

Let me know what you think and stay tuned for our next images titled, “ExcaLESbur” and “Lava Lamp!”