Our schedules finally converged. I caught up with Tom Bradicich, PhD, known within Hewlett Packard Enterprise (HPE) as “Dr. Tom,” to learn the latest on the converged edge. Tom is one of the half-dozen or so people I know who can dump so much information on my brain that it takes some time to digest and organize it. He led development of the Edgeline device connecting with the Industrial Internet of Things. He is now VP and HPE Fellow leading HPE Labs developing software to come to grips with the complexities of the converged edge and “Converged Edge-as-a-Service”.
He likes to organize his thoughts in numerical groups. I’m going to discuss converged edge below with his groupings:
3 C’s
4 Stages of the Edge
7 Reasons for IoT and the Edge
3 Act Play
12 Challenges
The foundation of the converged edge is found in the 3 C’s:
- Perpetual Connectivity
- Pervasive Computing
- Precision Controls
I remember Tony Perkins following up the demise of Red Herring magazine (charting the hot startup and M&A craze of the 90s, the magazine grew so large it came in two volumes for a while) with an online group called AlwaysOn. Trouble is, back in the 90s, we weren’t “always on.” Persistent connectivity was beyond our technology back then. Now, however, things have changed. We have so much networking, with more to come, that perpetual connectivity is not only possible, but also mundane.
HPE didn’t take a personal computer and package it for the edge. It developed Edgeline with the power of its enterprise compute along with enterprise grade stacks. It is powerful.
Then we have the 4 Stages of the Edge:
- Things—sensors and actuators
- Data Capture & Controls
- Edge IT (networking, compute, storage)
- Remote Cloud or Data Center
This is where Internet of Things meets the Enterprise.
Why do we need edge compute and not just IoT-to-Cloud? 7 Reasons:
- Minimize Latency
- Reduce bandwidth
- Lower cost
- Reduce threats
- Avoid duplication
- Improve reliability
- Maintain compliance
The Converged Edge is a 3-Act Play:
- Edgeline systems & software; stack identicality
- Converged embedded PXI and OT Link
- Converged Edge-as-a-Service
At this point in time, we are faced with 12 challenges to implementation:
- Limited bandwidth
- Smaller footprint for control plane and container
- Limited to no IT skills at the edge
- Higher ratio of control systems for compute/storage nodes
- Provisioning & lifecycle management of OT systems and IoT devices
- OT applications are primarily “stateful”, cloud unfriendly
- Data from analog world & industrial protocols
- Unreliable connectivity—autonomous disconnect operation
- Higher security vulnerabilities
- Hostile and unfamiliar physical environments and locations
- Long-tail hardware and software revenue model—many sites, fewer systems
- Deep domain expertise needed for the many unique edges
Of course, we could go into each of these items. Dr. Tom does in one of his latest talks (I believe it was at Hannover). We should pause at number 12, though. This is an often-overlooked necessity by AI evangelists and other predictive maintenance would-be disrupters. When you begin messing with industrial, whether process or discrete manufacturing, it really pays to know the process deeply.
I can’t believe I summarized this in less than a 600-word essay (is that still the common university requirement?). It is just an outline, but it should reveal where HPE has been and where it is going. I think its power will be disruptive to industrial architectures.
Trackbacks/Pingbacks