Enterprises collecting and analyzing data from their internet-of-things efforts have a choice: Transport it all...
back to their data center, move it out to the cloud or do more analysis out at the edge.
For the latter, a new set of hardware and software choices is emerging to bring the known benefits of virtual machines out into the internet of things (IoT), such as increased efficiency, ease of management and overall hardware reduction.
More companies (34%) said they analyze IoT data in their company-owned or leased data center facilities, according to 451 Research's latest survey, Voice of the Enterprise: Internet of Things, released this week. Another 23% of IoT data is analyzed on IT infrastructure located where the IoT data is generated, while 11% is done out in public cloud infrastructure -- both infrastructure as a service and platform as a service.
Survey respondents indicated they'll analyze more data out of the cloud (17%) in the next two years, while in-house and edge data analysis will dip to 31% and 19%, respectively.
"There's a bunch of arguments for computation at the edge with centralized orchestration," said Christian Renaud, a research director in 451 Research's internet of things practice. Most operational technology (OT) systems from makers such as Siemens AG, Schneider Electric SE, Hitachi Ltd. and NTT Data Corp. are already edge-based, because there wasn't any cloud computing when many of those systems were installed 20 to 30 years ago.
"Orchestrated compute, storage and networking at the edge is the first merger of OT and IT," he said.
To address these concerns, vendors are offering hardware with VMs within a ruggedized, hardened converged system, hoping to put tools available in the data center alongside IoT's things in just about every location imaginable.
Living on the edge
Tools to quickly ingest IoT data and get it to the cloud have gained ground in recent years, but for many IT shops, this transition creates worries about data-transfer cost, security, latency and possible data corruption, among others.
One alternative: Apply a compute layer at the edge to do at least some big data analytics.
David KingCEO, FogHorn Systems
"There is still massive room for improvement of efficiency and productivity," said David King, CEO of FogHorn Systems, based in Mountain View, Calif., which makes edge intelligence software. "You can't do it from the cloud; you have to do it on-prem, as close to the machine data as you can."
Take, for example, a large enterprise with a factory in Florida. The assembly line produces products worth from $10,000 to $100,000, with about 8% coming out defective. The raw material that's part of those defective items can't be reused, not to mention the loss of revenue because the defective item can't be sold.
FogHorn Systems' Lightning software platform, unveiled two weeks ago, runs complex event processing on servers outside the data center and alongside the production line to identify potentially defective items before it is too late.
"We can alert the operators that there is a high-percentage chance that, at the end, these products will not make it through quality control," King said.
Those materials can then be reused before an irreversible action, such as heating or painting, happens. Already, the software has reduced the defective rate from 8% to 4%.
Taking VMs out to IoT
Software such as FogHorn's is just one of the applications that may have to run at the edge, making use of the value of VMs on an oil rig, for example, in the same way as in the data center.
Hewlett Packard Enterprise (HPE) is trying to put more compute on a gateway to avoid aforementioned concerns about latency, cost and privacy, which give enterprises pause to send data to the cloud.
The company's Edgeline devices combine Xeon processors with storage up to 30 TB and systems management similar to what is found in the data center -- and now, they can run virtualized applications on all major hypervisors, including VMware vSphere, Citrix XenServer, KVM and Microsoft Hyper-V.
"Those are things we know, but we are shifting them left, to the edge outside the data center," said Tom Bradicich, vice president and general manager of servers and IoT systems at HPE. "It is more valuable at the edge because there is not a lot of room, not a lot of power and not a lot of skills."
In addition to the compute, storage and networking found traditionally in the data center, the Edgeline devices, which HPE calls Converged IoT Systems, include the ability to capture data and convert analog data coming from things such as turbines, windmills, pumps, valves and buildings into digital data. Such systems could replace what used to be 1,000 gateways sending data back to VMs in a colocation data center, Renaud said.
In recent years, IoT gateways have largely focused on connectivity and analytics, Renaud said.
"This is taking it much further than that, it has more than an x86 processor with a router card," he said.
Robert Gates covers data centers, data center strategies, server technologies, converged and hyper-converged infrastructure and open source operating systems for SearchDataCenter. Follow him on Twitter @RBGatesTT or email him at firstname.lastname@example.org.
Edge computing gets push from quality, IoT device concerns
IoT data collection, transmission needs edge network
Get your data center ready for the IoT era