SmartRural aims to help farmers worldwide use analytics to shrink their costs and ecological footprint. As a lean start-up, the company needed a more cost-effective option than scaling out its on-premise analytics platform.
SmartRural migrated its cutting-edge analytics offering to bare metal servers in the IBM Cloud, delivering high levels of performance, security and scalability without up-front capital investment.
Up to 60%faster analytics processing helps deliver rapid insights to farmers
EnablesSmartRural to grow its client base cost-effectively
Supportsinternational expansion with practically unlimited scalability
Business challenge story
Tackling the challenges of a changing climate
Climate change is accelerating fast. The Intergovernmental Panel on Climate Change (IPCC) forecasts a temperature rise of 2.5 to 10 degrees Fahrenheit over the next century, and agricultural producers around the world are already starting to feel the effects. As droughts and heat waves become more frequent and intense, the cost of farming is on the rise.
To continue to deliver the highest possible yields while minimizing their ecological footprint, it’s crucial for farmers to optimize their water use and protect crops from ongoing threats such as pests and disease. Enter SmartRural, a precision agriculture start-up based in Spain.
Sergio Rodriguez, CEO and Co-Founder at SmartRural, begins: “In recent years, Spain has been experiencing prolonged, severe droughts, which have hit our farmers and wine-producers hard. Three years ago, we saw an opportunity to harness the latest advances in drone photography, remote sensing and analytics to help farmers make better-informed decisions about everything from irrigation and fertilization to pesticide treatments.”
SmartRural built a location intelligence platform to combine its clients’ data with its own thermal, multi-spectral and RGB drone images of fields, as well as data from remote sensors embedded in the soil. By enriching this data with additional information from satellite imaging and meteorological services, the company can identify if water is being wasted in certain areas of a field, determine the fertility of the soil and health of the crops, and pinpoint areas of frost or insect damage.
Thanks to this location intelligence platform, and the tools that SmartRural uses to gather its data, the company can train highly accurate artificial intelligence (AI) models to give its clients insights, predictions and recommendations about how to efficiently manage their crops and avoid diseases.
Ana Antunes, Agronomist at SmartRural, comments: “We collect large amounts of data for each of our clients, and to process it effectively we rely on analytics models and machine learning. For example, when we capture an RGB image of a field, a member of our team first trains the model to recognize areas of insect damage, after which our analytics software completes the process automatically.”
In the past, SmartRural utilized a commodity x86-processor-based environment to support its analytics requirements.
“Our analytics workloads are computationally intensive,” says Antunes. “In many cases, we need to utilize 100 percent of our CPU and GPU resources and up to 500 GB of RAM to process just one client’s data. Because our resources were limited, completing an analysis took many days on our in-house server platform, which reduced the number of clients we could work with simultaneously.”
As the company engaged new clients across Spain and targeted international growth, it realized that scaling out its infrastructure in line with the fast-increasing demand for compute, storage and networking resources would drive up costs substantially. To continue its international expansion without sending costs soaring, the company decided to migrate to a cloud platform.
Delivering precision agriculture insights in the IBM Cloud
To meet the demanding requirements of its workloads, SmartRural selected two dedicated bare metal servers in an IBM Cloud data center in the Netherlands. The bare metal platform offers SmartRural high levels of performance, security and scalability without the need for up-front capital investment.
Pablo Aibar, Data Scientist at SmartRural, says: “To complete our workloads quickly, we need to fully utilize our CPU and GPU resources for weeks at a time—and we felt that a bare metal environment was the optimal choice to achieve our utilization target.”
The company uses a primary server running Microsoft Windows Server 2016 to run its proprietary image-stitching software and other demanding processes such as deep learning models. To optimize the server for highly parallelized image processing, machine learning and computer vision training tasks, SmartRural deployed two NVIDIA Tesla P100 GPU accelerators.
SmartRural’s image stitching and deep learning processes often require up to three times their disk size in RAM—and because the source data can total up to 300 GB, SmartRural provisioned 1 TB of RAM for its primary server.
The company uses its second server to drive less intensive data visualization and reporting tasks, which run in Docker containers on Debian GNU/Linux. As these use cases have lower in-memory processing requirements, the company configured this server with 256 GB of RAM.
SmartRural received generous support from IBM to set up and migrate to the new platform. Antunes comments: “One of the things we appreciate most about IBM is the key role they play in incubating and supporting start-ups. Through the IBM Global Entrepreneur Program, IBM offered us the ability to migrate our on-premise environment to an enterprise-class platform in the IBM Cloud with no up-front costs—an invaluable advantage at a critical stage of our business growth.
“We completed our cloud migration in a matter of weeks. As a start-up, advice and support from global businesses like IBM is a real asset—and on that front, our partnership with IBM is a major advantage to achieve our long-term vision.”
Like all organizations that store personal data on their clients’ behalf, SmartRural must ensure its compliance with requirements such as the General Data Protection Regulation (GDPR). By moving to a bare metal environment in the IBM Cloud, SmartRural has the control and visibility it needs to demonstrate compliance.
Aibar comments: “Currently, most of our GDPR-regulated data is managed outside the cloud—and we encrypt the sensitive data we do hold in a partially isolated container on our bare metal servers. We always welcome additional ways to make our environment more secure, and we are interested in exploring solutions such as Intel Trusted Execution Technology [TXT] and HyTrust for our IBM Cloud platform.”
Boosting yields, cutting costs and protecting the environment
With the IBM Cloud supporting its precision-agriculture offering, SmartRural can analyze more data for more clients faster than ever before—offering the headroom it needs to realize international growth.
For example, SmartRural has achieved a massive increase in the number of wine growers it can serve. Previously, the process of training SmartRural’s analytics models to count the total number grapes on vines to estimate yield, assess the maturity level of crops and detect issues with fungus and pests was a time-consuming process, which put a hard limit on the number of clients the company could service in a season. By migrating to a parallelized platform in the IBM Cloud, the company now harnesses high-performance GPUs to accelerate these vital workloads.
Aibar explains: “Computer vision processing is a key enabler of our services, as it allows us to automate essential but time-consuming data-collection processes. Utilizing the parallel-processing capabilities of GPUs in our bare metal environment allows us to train our computer vision models 15 percent faster, and accelerate our data processing by up to 60 percent. This is enabling us to deliver results to farmers sooner, and increase the number of clients we can work with in any given season.”
SmartRural is also measuring dramatic improvements in processing performance for raw data from its remote sensors in the soil, which offer valuable insights into the fertility of each field.
Antunes continues: “We can now crunch sensor data as much as 10 percent faster than before. By analyzing farmers’ data more efficiently, we can support more of them with the actionable insights they need to deliver good crop yields cost-effectively in a changing climate. In fact, we help our clients to reduce their irrigation, fertilizer and phytosanitary costs by an average of 25 percent.”
Based on its successes with the IBM Cloud, SmartRural is already looking to the future. The company is interested in integrating meteorological data from The Weather Company, an IBM Business, to augment its existing data sources with more accurate, granular insights. The company also plans to explore the AI capabilities of IBM Watson® to train its analytics models, which would enable it to cut time-to-insight further still.
Antunes concludes: “As our climate changes, we need smarter ways to make the most efficient use of the world’s limited resources. Thanks to the IBM Cloud, we have a future-ready platform to help farmers around the world use precision agriculture to enhance the efficiency of their production.”
Based in Valladolid, Spain, SmartRural supports farmers with technology solutions to enable cost-effective, environmentally friendly operations. Combining cutting-edge geological investigation, drone photography and analytics methodologies, SmartRural helps clients to design more effective irrigation systems, combat pests and increase crop yields.