How to choose an edge gateway

There could be as many as 15 billion IoT devices connected to enterprise infrastructure by 2029, according to Gartner. These devices will generate massive amounts of operational data that needs to be translated from their original protocols, aggregated, and analyzed in order to deliver real-time actionable alerts as well as longer-term business insights.

For organizations with significant IoT deployments, edge computing has emerged as an effective way to process sensor data closest to where it is created. Edge computing reduces the latency associated with moving data from a remote location to a centralized data center or to the cloud for analysis, slashes WAN bandwidth costs, and addresses security, data-privacy, and data-autonomy issues.

[Get regularly scheduled insights by signing up for Network World newsletters.]

On a more strategic level, edge computing fits into a private-cloud/public-cloud/multi-cloud architecture designed to enable new digital business opportunities.

One big challenge of edge computing is figuring out what to do with all the different kinds of data being generated there. Some of the data is simply not relevant or important (temperature readings on a motor that is not overheating). Other data can be handled at the edge; this type of intermediate processing would be specific to that node and would be of a more pressing nature (the motor is overheating). The cloud is where organizations would apply AI and machine learning to large data sets in order to spot trends (if we adjust our maintenance schedule, we can prevent all motors from overheating.)

The fulcrum that balances the weight of raw data generated by OT-based sensors, actuators, and controllers with the IT requirement that only essential data be transmitted to the cloud is the edge gateway.

The difficulty that enterprises face is that the market for edge gateways is relatively immature. In addition, edge gateways aren’t one-off purchases; they have to seamlessly integrate with enterprise-owned infrastructure such as IoT devices and with third-party resources such as cloud-provider analytics service.

The good news for IT execs looking to deploy edge gateway technology is that the vendors you already have longstanding relationships with are offering edge gateways, either as hardware or as a managed service. The list includes Cisco, HPE, Dell, IBM, Microsoft and AWS.

What to look for in an edge gateway?

Here are some of the key questions to ask when narrowing down your search for an edge gateway.

Is the edge gateway compatible with the application- and networking-layer protocols running in my infrastructure or might be in the future?

Application layer protocols include MQTT (Message Queuing Telemetry Transport), ZeroMQ, CoAP (Constrained Application Protocol), AMQP (Advanced Message Queuing Protocol) and DDS (Distributed Data Service). Networking protocols include Wi-Fi, Zigbee, Bluetooth, Bluetooth Low Energy (BLE), Z-Wave and LoRaWAN. And 5G is certainly looming. Then there are network protocols specific to industrial automation such as Modbus, BACnet, CANbus and LonWorks.

The edge gateway needs to have the ability to speak these languages, integrate all of that data, and draw a coherent picture of what’s happening on the factory floor, in the oil field, or across the smart city.

Does the edge gateway support tight security, encryption, and Zero Trust?

Gartner warns that “edge computing workloads and systems will be targets for attack like any other IT systems, with the added complexities of physical insecurity, large number of attack surfaces, nascent end-to-end security solutions, intermittent network connectivity and minimally equipped endpoint devices.”

Look for an edge gateway that provides data-at-rest encryption, boot time integrity checks and identity management that covers both hardware and software. Edge gateways should be included in any deployment of Zero Trust, which provides stringent authentication authorization, or secure access service edge (SASE), which blends SD-WAN with security.

Is the edge gateway ruggedized?

Putting compute resources at the edge provides many benefits, but the gateways need to be designed to handle harsh environments, which includes extreme heat or cold, vibration, pressure, dust, dirt, moisture, etc.  And the gateway needs to function in situations where cloud connectivity is limited or intermittent. Depending on the deployment scenario, you might also want to ask whether the gateway has a wall-mounted version.

Is the gateway built on open standards?

Given the nascent nature of edge deployments in general and the fact that the gateway needs to talk to all of the other components of the edge model, it’s probably a good thing to keep as many options open as possible, which means open standards.

For example, one framework that is gaining acceptance is the Linux Foundation’s LF Edge, an umbrella organization that includes multiple open-source, edge-related projects including EdgeX Foundry (based on code that Dell originally wrote internally and then open sourced and donated to the Linux Foundation) and Akraino Edge Stack. These software packages, many of which are still in the development stages, are designed to eventually provide an open-source architecture for the edge.

What is the management system for the edge gateway?

If you’re putting edge gateways in all of your manufacturing plants across the world, for example, then you’ll need to be able to manage them securely, remotely and efficiently. Look for vendors who offer tight integration between the gateway itself and a cloud-based management dashboard.

Hardware options for edge gateways

There are others, but three of the most familiar names in networking are offering edge gateways:

HPE

HPE Edgeline edge gateways are built on the same technology as HPE’s data center systems and are designed on four principles: open, standards-based architecture, ruggedized form factor, security and manageability, and smooth integration between OT (control systems, data acquisition systems and industrial networks) and enterprise IT. 

The HPE edge gateways run on Intel Xeon processors or GPU accelerators, and there are different models, depending on compute, memory and storage requirements. Gateways can also be configured with HPE server blades to provide higher capacity processing. The systems support industry-specific protocols such as Modbus, BACnet and OCP-UA.

Dell

The Dell Edge Gateway 5000 series for IoT is targeted at building and industrial automation verticals. It features an Intel Atom processor, solid-state hard drive, Gig Ethernet connectivity, an optional CANbus card and support for Wi-Fi, BLE, Zigbee/LowRaWan. It’s ruggedized and can be deployed on a wall or DIN rail at the edge of the network. According to Dell, its edge gateway “has the capacity to perform analytics locally, close to the devices and sensors generating data, sending only meaningful data to the cloud or data center.”

For security, the Dell gateway features a Trusted Platform Module (TPM) chip for hardware root of trust, secure boot and BIOS-level lockdown of unused I/O ports. Management is provided by Dell Edge Device Manager software.

Cisco

Cisco IoT Gateways, part of a broader line of edge routers, are designed for easy installation at scale through remote, automated provisioning via Cisco Control Center. Gateways are managed centrally with Cisco’s cloud-based IoT Operations Dashboard. There are ruggedized models, and Cisco security extends from the edge to the headend in the data center or to the cloud.

There are also other vendors selling edge gateways, including Advantech, NEXCOM, Eurotech, Kontron and Adlink.

The purchase of an edge gateway implies a more strategic decision to build out your own edge platform, independent of the hyperscale cloud that it will ultimately connect with. Organizations that go this route are more likely to have their own application development environment, and have the resources, skills, and time to pull it off. This approach avoids cloud vendor lock-in, but it does have its risks.

Gartner cautions that “one of the greatest impediments to deploying edge computing is the lack of broadly accepted infrastructure and operation models, leaving the enterprise to piece together a solution from still emerging technology stacks and operational models.”

The alternative approach is to take advantage of fully managed edge services provided by the hyperscalers.  

AWS, Azure, IBM bring cloud analytics to the edge

By 2023, 70% of enterprises will run some level of data processing at the IoT edge, and more than 50% of new enterprise IT infrastructure will be deployed at the edge rather than corporate data centers. Not all of it will be purchased outright, however: More than 75% of infrastructure in edge locations will be consumed and operated via an as-a-service model, according to IDC.

For organizations that want to go that route, cloud services providers like Amazon, Microsoft, and IBM will happily deploy edge gateways for you, as long as your data processing takes place in their cloud.

AWS

AWS is extending its portfolio of cloud-based services, APIs and tools from the cloud to the edge as a fully managed service. High-performance apps built in the cloud that require data processing and storage can be deployed at the edge to deliver “ultra-low latency, intelligence real-time responsiveness, and reduce the amount of data transfer.”

AWS has a laundry list of edge software and services, including FreeRTOS, an open-source OS for microcontrollers, IoT Greengrass, which extends AWS to edge devices, AWS IoT Device Defender, IoT Analytics, IoT Events and IoT Device Management. SiteWise is the AWS edge gateway.

In addition, AWS can provide the hardware: AWS Outposts is a compute and storage rack that is delivered fully assembled and is installed by AWS technicians. AWS also has a line of Snow Family edge devices aimed at locations where there is a lack of network connectivity.

Azure

Similarly, Microsoft’s Azure IoT Edge is a fully managed service built on Azure IoT Hub, a cloud-hosted service that enables organizations to connect, monitor and manage IoT assets at scale. IoT Hub delivers per-device authentication, built-in device management and scaled provisioning. The IoT Edge service is designed to “process the data locally and send only what’s needed to the cloud for further analysis.”

On the hardware side, modules are designed to run as containers on enterprise servers (Windows or Linux). Microsoft has also teamed with third-party vendors to build a list of ‘Azure certified’ devices, such as the ReliaGATE edge gateway.

IBM

IBM comes at edge computing from a unique perspective because it is both a hardware vendor and a cloud services provider. The IBM approach is to offer a fully managed service that provides tight integration between the physical edge gateway/servers and the IBM Cloud, which has its own powerful analytics capabilities (Watson).

IBM is focused on use cases involving sensors that generate large amount of data, which requires heavy processing at the source. These include factory floor, telecommunications, construction, utilities, retail, industrial automation, content production and gaming.

When edge matters most

Multiple trends are converging to make the edge a critical part of any enterprise IT architecture. These include 5G, which will enable new mobile applications; the rise of AI, which will make it easier to process data at the edge; and the fact that data is moving inexorably from a central core to the edge.

According to the Linux Foundation’s 2021 State of the Edge Report, the pandemic accelerated this shift because companies needed to conduct remote monitoring and management of multiple edge nodes–work-from-home employees.

“Our analysis shows demand for edge infrastructure accelerating in a post COVID-19 world,” says Matt Trifiro, co-chair of the State of the Edge project and CMO of Vapor IO. “The new digital norms created in response to the pandemic will be permanent. This will intensify the deployment of new technologies like wireless 5G and autonomous vehicles, but will impact nearly every sector of the economy, from industrial manufacturing to health care.”

For companies just starting on their IoT edge journey, Gartner recommends these steps: Appoint an edge architect or edge architecture committee, look for current applications that could benefit from the advantages that edge computing can provide, and then look for opportunities to deploy new applications at the edge that provide business benefit.