Augury uses Google Cloud big data technology including Cloud Dataflow and BigQuery to push data from basic Cloud Storage buckets into BigQuery tables, which enables tens of millions of machine learning features. Being able to quickly and easily load data without complicated ETL processes allows researchers to get quick insights while running fast queries. It also enables Augury to run faster research cycles and enhance the algorithms that predict machine failures.
Improving the flow of ideas between research, development, and production
Both DoiT and Augury appreciate that Google Cloud offers a broad range of IoT and research-oriented products, releasing new products frequently that not only enable research and development, but remove the need for the company to build tools in house. Developing basic functionality such as registration, authentication, and data transfer wasn’t where Augury wanted to spend its resources.
Google Cloud products such as Cloud IoT Core, Google Kubernetes Engine, Google Cloud Dataflow, Cloud Dataproc, and Cloud Datastore enable Augury to use the same data for their research and production environments, which Gal finds extremely helpful. He explains that being able to see the growth and rapid iterations that come from working with open source pipeline technologies such as Google Cloud Dataflow and Apache Beam allows the Augury team to work on algorithms with a proof of concept, agile mindset, while also contributing back to the open source community. Working in pipelines allows Augury to move back and forth from research to production quickly, enabling the company to deliver value as fast as possible to its customers. This is a new concept for factories, which are experiencing true digital transformation with Augury solutions.
“We start with an experiment,” says Gal, “if it works well, we can expand it rapidly, and if it doesn’t, we can try something else quickly. Google Cloud gives us the tools to understand what’s right for us, for our customers, and for our ever-changing industry.”
Being able to run in multiple environments was also a selling point of GKE. Though Augury runs on Google Cloud, some of its customers use other cloud solutions. With customers sometimes having different needs, the ability to migrate microservices from one place to another while remaining always available was extremely important to the company.
Delivering business impact through machine health insights
Augury is able to keep its promise of delivering continuous insights to customers by relying on the direct capabilities of Google Cloud IoT technologies. Google Cloud IoT Core, Cloud Pub/Sub, and Google Cloud Dataflow make it easy for Augury to consume telemetry from IoT devices deployed on factory floors. This constant flow of incoming data allows for ongoing improvement of capabilities to monitor and fix problems with IoT devices, thanks to early detection of connectivity issues. This makes sure that sensor data arrives in an uninterrupted manner without data loss and that customers receive early warning of developing machine failures.
Following rapid growth with its customers, Augury faced initial challenges in managing the sheer volume of data coming in through its IoT devices. The autoscaling features included in GKE solved this by ensuring the solid and balanced algorithm processing latency that meets the Service Level Agreement (SLA) that Augury has with its customers. Even if a large facility comes back online after having been offline for a long period, pushing huge volumes of data in the process, GKE can handle the flow. With GKE autoscaling, the company can control the number of algorithm instances running in response to data flows and keep task queues low.
Augury has been able to grow its field deployments and improve its diagnostics capabilities without affecting service performance since adopting GKE. Thanks to the management layer for monitoring services such as memory consumption, GKE allows Augury to identify and fix issues early in order to avoid problems that can cause performance issues to other parts of its system.
This emphasis on continuous data flows and analysis means that Augury customers get the benefit of real-time insights into machine health. Some manufacturing customers look at this data as often as hourly, and rely on Augury to consume, analyze, and deliver insights as quickly as possible. GKE plays an important role by ensuring that Augury instances will always run in a balanced manner, allowing the company to provide seamless service and insights to customers.