Ivelin Radkov - Fotolia
A full story of the HPE IoT strategy takes a good number of sources from different acquired companies to assemble. In this, the world's largest data center vendor is no different from its largest competitor: Dell EMC. Both vendors also name the same drivers for their enterprise-site gateways and IoT-specific computers: the need to move data analysis and response closer to "things" and their sensors, and to the controllers that make things such as pumps, valves, doors and robotic arms move in response to sensed conditions. Just as our reflexes fire muscles to jerk our hands from fire without the wasted time needed to send signals all the way to the brain, many automated responses in industrial environments must take place too quickly to tolerate the latency of the cloud. In addition to the urgency for speed, other drivers are bandwidth economy and security.
Hyper-convergence is another common theme. In Dell EMC's case, this means systems combining compute, storage, hybrid cloud orchestration and connectivity. In HPE's, it's the combination of connectivity, compute power and an inclusion that Tom Bradicich, HPE's VP and general manager of servers, converged edge and IoT systems, said is new: control systems. The outcome of an alliance between HPE and National Instruments, HPE is first to build the physical response signaling into its converged IoT systems, branded Edgeline. Not coincidentally, National Instruments is Bradicich's immediate former employer.
HPE's Edgeline series, introduced at its London Discover 2016 conference, begins with the EL10 Intelligent Gateway and, like Dell EMC's, it's ruggedized and fanless for industrial environments. With dual-core Intel Atom CPU, 4 GB RAM, 32 GB SSD storage and Wi-Fi connectivity, it's for lightweight sensor data aggregation, acquisition and analytics. Next up, the midlevel EL20 comes with an lntel i5 CPU, doubling RAM and storage.
HPE IoT focuses on shifting computing power to the 'thing' side
Starting with the EL1000 and up, HPE is moving data-center-scale compute power into the edge -- "shifting left," to where the "things" are usually represented in diagrams, according to Bradicich. Here you have up to 64 Xeon processors and up to 40 TB of storage -- enough for extremely computationally intensive tasks like image analysis and breakdown forecasting.
The pieces of the HPE IoT connectivity and security puzzle partly come out of its 2015 acquisition of Aruba. The switch maker's "IoT-ready" 2540, also announced at Discover, now addresses security through its ClearPass Policy Manager, an asset Aruba acquired with Avenda Systems in 2011. This software classifies and assigns network access policies to all mobile and IoT devices, both wired and wireless, gating user access privileges based on device type, ownership status, operating system or user-defined variables. ClearPass also runs on the EL10 and 20.
As Trent Fierro, Avenda alum and Aruba's director of security marketing explains it, ClearPass performs rule-directed, automated tunneling that walls off devices and their servers from the rest of the network by function. Or by bandwidth: An HVAC sensor that suddenly started using 10 MBps would be bounced off the network. According to Fierro, this would prevent breaches like the infamous Target hack, where the company's POS systems and customer data were reached through the HVAC system.
Come for Universal Profiler, stay for ClearPass
Universal Profiler, a subset of ClearPass, is being introduced as a stand-alone introductory product. It uses DHCP, SNMP, Nmap scans, NAD discovery protocols and a variety of other means to inventory and "fingerprint" what's on your network. Without this information, Fierro said, you can't know your risk of exposure. "For example, those CCTV cameras that were being connected by that company that was shipping them out of China. They had a preset login and password that you couldn't change." Someone who knew that logged in and started the Dyn DDoS attack that took down CNN, Reddit, Netflix and other popular sites in October.
ClearPass Policy Manager's APIs integrate with third-party products to attach other kinds of criteria to network access control. Fierro mentioned a Check Point firewall as an example, checking for such suspicious user behavior as 10 login attempts in 10 minutes.
Another new IoT piece, HPE's Universal IoT Platform, is a series of software modules for on premises or cloud that receives sensor data, normalizes it to HPE's oneM2M data model and presents it to applications for analysis. It may enrich the data with other information, like weather or stock prices.
HPE IoT demo on the London stage
To help introduce the pieces and partners of its IoT ecosystem, HPE's Tom Bradicich and Howard Heppelmann, general manager of connected manufacturing at PTC, conducted a demo at the London Discover event. They wheeled in a working station of an automated industrial assembly line, where a robotic arm continuously picked up components from one conveyor belt and placed them on another. Sensors on the whole real-life installation, for an unnamed food packaging facility, generate 800 data points per second, Heppelmann said, all digested by the HPE Edgeline 1000 running PTC's ThingWorx IoT stack. In Heppelmann's hands, an iPad offered three ThingWorx applications, each aimed at a different operations role: operator, plant manager and maintenance engineer.
Factory performance -- typically based on static or outdated data -- was captured in the demo in real time to accomplish three things: first, to continuously sharpen the accuracy of tolerance ranges and second, based on those ranges, to detect anomalous sensor readings -- for demo purposes, air flow and pressure. The third aim: predicting future problems likely to be caused by these anomalous conditions and -- most interesting -- the time window in which they are most likely to occur.
Choosing the maintenance engineer's app, Heppelmann showed his screen listing the health status of devices for which that role held responsibility. He then asked Bradicich to open a valve, introducing an air leak into the station's pneumatic system. The air flow sensor detected the leak and alerted the PTC app on the iPad. Further, ThingWorx's machine learning engine predicted a pneumatic gate failure in 10 days, and automatically generated a maintenance work order for its repair. With its prognostic analysis at the edge, the system achieved the Goldilocks of predictive maintenance; not too much, and not too little, too late.
Heppelmann then took the demo a step further to a different mechanical problem, affecting a conveyor belt. Here ThingWorx superimposed animated graphics of repair steps, augmented reality-style, over a live iPad-camera view of the unit under repair -- like a printer shows you how to fix a paper jam, only better, because the user trains the iPad's camera on the live "patient," while PTC's ThingWorx guides him through the process in 3D.
HPE's demo illustrated the impracticality of cloud-based computing in this scenario; the cloud could not manage the real-time response needed on factory floors. Like other vendors, though, Bradicich sees an IoT role for the cloud in longer-range analysis, and in aggregating and archiving the IoT activity of several sites.
HPE IoT vs. Dell EMC IoT
Analysts, asked to compare HPE IoT offerings and approaches to Dell EMC's, are quicker to point to market messaging, partner evolution and emphasis than core technology. Maribel Lopez, an analyst at Lopez Research, finds HPE somewhat behind Dell in simply crystallizing its IoT story. "HPE has a very compelling big data and analytics portfolio," she said, but having lost some mindshare to its big spinoff stories of 2015, it's not as far along in explaining its IoT ecosystem to the marketplace.
Mike Krell, IoT analyst at Moor Insights & Strategy, said HPE has a "leg up" in the carrier space, essential for connecting widely dispersed things to the cloud or data center. Also unveiled in London Discover, HPE's Mobile Virtual Network Enabler (MVNE) product aims to help carriers bring cellular costs down way below those of phones, to something practical in very large-scale deployments. In its release, HPE said that MVNE's provisioning, configuration, administration and billing services would give rise to IoT-specific mobile virtual network operators.
Krell also credited the vendor with "a strong play in its home-developed Universal oneM2M platform, and I see it trying to leverage that into many situations." At the same time, he echoed Lopez when he noted, "I haven't really seen an IoT strategy from HPE that encompasses the entire HPE offering. I have seen a number of presentations from Dr. Tom [Bradicich] where he doesn't even mention its oneM2M platform." This now goes by the name Universal IoT platform, one sign that HPE is indeed working on integrating the marketing efforts of its constituent IoT companies.