Delivering Faster Insights via Edge Computing

Whether you are looking to achieve full autonomy or simply looking for ways to reduce your data costs at the cloud or historian level - Edge Computing (or "Edge" for short) can help. Below is the quick and dirty on how.

What is Edge Computing, Really?

Let’s use the analogy of watching a movie to describe edge computing. Nowadays, you can watch movies in several different ways –- a few examples being Netflix, Hulu, and good ol’ fashioned movie theaters. But edge computing is Blockbuster. Or better put, the mechanism that allowed us to watch Blockbuster movies locally, a VCR player, is the edge device. Where does "Intelligent Edge" lie? Think of this as downloading a movie onto your iPad before a flight- the iPad being the edge device.

Figure 1: What is Edge Computing? Source: Real-Life Use Cases for Edge Computing, IEEE Innovation at Work

Why Go Edge?

Instead of being at the mercy of the big boys’ cloud compute power (i.e. AWS, Microsoft, Google, and IBM), edge computing allows you to take back some control and achieve 1. Better Speed, 2. Top-Notch Privacy & Security, all while 3. Managing Bandwidth Constraints.

We all love the convenience of centralized (cloud) computing in our day to day lives, (So you’re saying I DON’T have to get in a car, drive to the Blockbuster, create a membership, browse the aisles, have a debate on Horror vs Rom Com, choose a movie, choose another movie, checkout, and drive home?) but when it comes to certain applications in industry, it’s important to consider whether things like latency, privacy & security, and/or bandwidth will be a critical component of the specific use case you are considering.

Where Edge DOESN'T Make Sense.

Once when visiting NYC, my Texan family had the wonderful idea to have an 8 pm dinner on a Saturday night and DRIVE from Princeton, NJ to the Times Square area. One car. Four hangry individuals. And 3 cell phones, 1 on Apple Maps and 2 on Google Maps, searching for parking. Long Story Short - It took 1 hour for the nav apps to route us to NYC and 1.5 hours to find parking. Can you guess what the debate between the three cell phone holders was? (Spoiler Alert: Whether to turn left or right!)

In retrospect, it was clear that each instance of Apple and Google Maps apps were likely experiencing network issues, perhaps due to the sheer amount of individuals that also decided to drive near Times Square at 8 pm on a drizzly, Saturday night. By the way, we did not make the dinner reservation on time (in case you appreciate closure in a story).

There is a time and place to rely on cloud computing – other times, it may not make sense.

Where Edge MIGHT Make Sense.

A quick overview of a couple of use cases.

1- Learn in the Cloud, implement at the Edge – Similarly to the NYC story, for some oilfield applications, there is a window of opportunity where actionable insights are actually actionable and any delay in the implementation will be suboptimal. Running these control systems at the edge and allowing time-sensitive AI/ML models to be implemented locally, can be one way to achieve this. One thing to consider when looking at running your optimization techniques at the edge vs in the cloud, is the data’s frequency and resolution, which leads to the next potential use case for edge computing.
2- Data Quality Management at the Edge– Is it possible that in my “NYC parking at 8 pm” story, perhaps Apple/Google was ingesting large amounts of data and was having trouble contextualizing the data across the three different cell phones in a fast, repeatable manner? Definitely. More data can be messy. Especially when there's several different variables at play. But if the compute, storage and processing of data can be complete before sending it through the internet and in to the Cloud, then perhaps that not only will improve the quality of your actionable insights, but allow you to achieve some savings on the transmissions cost side.

Figure 2: Oilfield Benchmarking: How old is your data? Results from the Field Efficiency & Production Data Quality Forum.

Let’s talk Tech.

There’s a sea of “IoT” technology providers with the potential to work at the edge, majority of those relating to production technologies can be found on our Oil & Gas: Real-Time Data framework pictured below.

Figure 3: Tech Landscape for Real-Time Data for O&G Production

When we look at edge computing vs cloud computing, its gotten much more convenient to go the cloud route since as an operator, you can avoid spending capital on the hardware and infrastructure associated with edge compute and you do not have to deal with the logistics of training/managing the IT side of things. Even if you were willing to invest in edge compute on a large scale, the vendor side is flooded with fairly generic hardware, such as server racks, from the major incumbents (e.g. Hewlett Packard, IBM, DellEMC, Rockwell, etc.) with not many flexible options available, oftentimes being a pain to manage from the ground up.Then there's edge gateways which try to address some of the issues with traditional SCADA setups, but limits still exist since they don't have the storage and CPU embedded into the devices, making the full cost to implement often times pricier.

Companies like Hivecell are looking to shake things up by offering the compute power in a modularized, easy to install & maintain way, and also delivering some out of the box benefits, such as cloud replication & edge data processing, through their partnerships with Apache Kafka & others.

Hivecell Description

Link to profile.  

  Hivecell provides an edge computing platform as a service, through their Hivecell edge device which incorporates modularized, easy to configure, encrypted mini servers that do not require additional equipment, such as UPS or routers. They are looking to enhance your data pipeline and edge stream process to provide scalable, on-premises computing in a Platform-as-a-Service business model, alleviating any IT/training headaches.

Another tech provider worth mentioning who is not so much an edge computing provider, but more so has technology to enable edge-first operations that do not require traditional SCADA infrastructure, is WellAware who recently presented at the WellAware: Field Efficiency & Production Data Quality for Escondido Forum. Although WellAware technology will not have the out of the box capabilities to run deep analytics in the field, they can be seen as a competitor to the gateway space, and do have the capabilities to normalize your data closer to your edge, addressing some of those use cases where edge compute may make sense.

WellAware Description

Link to profile.     WellAware provides a full stack, edge first, IoT solution to monitor & control production equipment such as pump jacks, flowmeters and tanks. Their most developed application is on the chemical injection side. Communication to the cloud can be done via their Rest API, Secure FTP or MQTT.

This blog is meant to get the conversation started on a topic that historically was more of an IT issue, but with this convergence of IT and OT that many companies are experiencing, may be worth taking a second look at.

Let me know your thoughts on the topic below & comment if you think this would be worth a full forum.