#HPCMatters – Growing Your Food

This entry is part 1 of 7 in the series #HPCMatters

When I hear the word farming, the first thing that comes to mind is my maternal grandfather, Pawpaw Karnes, working on his farm. I think of hay bales, cattle, crops, and robbing the bees; I think of dirt, sweat, and the smell of country air, and I imagine things that have been true of farming for at least 150 years if not since the beginning of time. I, like most people, do not imagine datacenters and high performance computing. I, like most people, am leaving out an essential, ever-growing piece of what farming is today.

The Evolution of FarmingComputers in the wild

If you’re establishing a farm, you have to decide whether to plant crops or raise livestock, or how to mix your different options. At some point in time, a farmer might’ve asked neighbors, done some looking-around, and made a decision. Nowadays, there are simulators that will attempt to predict the 30-year output of a farm. These simulations take into account years of historical weather data, typical impacts of farming activities, likelihood of quality grazing, the cost of buying feed, soil nutrient levels over time, and virtually every major factor for crops or livestock yields. Many of these kind of simulations can be executed on a single desktop computer, and can factor more data into the decision than a farmer surveying the land.

Other kinds of simulations are done to inform all kinds of farming decisions. Among the many problems being solved by simulations and computational analysis are: comparing crop yields from year to year, studying root systems to determine better ways to grow crops in adverse conditions, and leveraging bioinformatics to increase crop yields. These advances are literally helping us feed the world more efficiently, and are constantly updated for changing conditions on continents and accounting for more and more variables such as the conditions of the soil.

The Cutting Edge: Informing Decisions In Near Real-Time

We have discussed many of the tools to help plan what crops to plant, how to update year-to-year, etc, but the cutting edge affects the day-to-day maintenance and decisions of farming. In the most advanced systems, sensors are everywhere across the farm. They can report anything about the health of the farm: moisture levels, current progress of a crop or seed-line, and so on. This information can be used to predict crop yields, compare to crop yields, and make decisions about whether any adjustments need to be made in how the farm is cared for.

In the natural process, there are very strict deadlines for certain kinds of analyses, especially during the time windows for planting and harvesting. Apart from that, there are many jobs to predict the next year, compare which line of seeds works the best in which environment, and integrating this data with all predictable variables that will affect the crops for future years. Mixing and prioritizing all of these jobs requires a policy-rich scheduler such as Moab to make sure that all hard deadlines can be met while making steady progress on jobs related to longer-term analysis and minimizing thrashing for the long-term workload.

The most modern farming systems current use HPC as described. In the future, I would expect more and more HPC jobs around farming. We know that the world will always be changing and for the foreseeable future population will increase. Farming continues to adapt, and the emerging challenges will likely be solved more and more in the datacenter.

Series NavigationThe UberCloud Experiment Shows Why #HPCMatters >>
Facebook Twitter Email

Speak Your Mind