Formerly known as Wikibon
Close this search box.

HPE’s Seven Factors Shaping Highly Effective IoT Architecture

Premise: Physics and policy matter in computing, but especially in “internet of things” (IoT) applications, which invoke the most stringent constraints ever faced by tech professionals and will require clear headed architectural decisions to perform.

IoT has captured the imagination of business and technology professionals, but imagination can run wild, especially in groups. Many imagine that data movement can be instantaneous, processing cost-free, and storage “stateless” — in both technical and geopolitical terms. While IoT should spur the imagination, promising significant advances in automation, energy efficiency, and human engagement systems, a practical “edge” has to be added to the thinking. Recently, at HPE Discover 2016, Wikibon analysts had a chance to interview HPE’s Dr. Tom Bradicich and National Instruments’ Eric Starkloff about IoT trends and issues. The catalyst for the conversation was HPE’s new family of Edgeline IoT systems, but the discussion centered on how:

  • IoT assimilates the real and digital worlds. IoT-related technologies are powerful and diffusing rapidly, but essential concepts regarding opportunities and limits are still evolving. HPE and National Instruments are partnering to provide tools, technologies — and concepts — necessary to drive new IoT opportunities.
  • IoT architecture will be shaped by 7 factors. Not all applications performing IoT-related services or employing IoT-related data will operate at the edge, but most will. Why? HPE identified 7 factors that favor edge-based architectures for most IoT solutions.
  • Information technology and operational technology are converging. IT and OT historically have been separate domains, in part because of the stringent real-time and security needs of OT. However, IoT is forcing the two sides to bridge gaps, ensuring that the analog world of OT converges with the digital world of IT.

IoT Assimilates the Real and Digital Worlds

IoT is more than its sensors and controllers and actuators, the cloud, and software parts. In important ways, IoT is about assimilating the real and digital worlds: making real objects programmable. Evolving digital business requires capturing and translating the analog data generated by people and things — sound, temperature, visual, words, thoughts, insights — into digital data that then can be moved, processed, and stored utilizing the wealth of available digital technology. The goal of IoT, then, is to instrument and actuate work and play, with tighter cycle times.

“The analog world is literally infinite. And it’s only limited by how many things we want to measure, and how fast we measure them. And the trend in technology is more measurement points and faster.” Eric Starkloff, National Instruments.

IoT Architecture Will Be Shaped By 7 Factors

Making possible the programming of real objects opens up enormous business possibilities, but IoT architectures will still be subject to inviolable physical — and pretty unassailable social — constraints. These constraints guarantee that significant portions of IoT work will happen proximate to the objects themselves:

“In the applications that <National Instruments> serve, you often need to make a decision in microseconds. And that means that the processing needs to be done . .  on the edge, at the thing itself.” Eric Starkloff, National Instruments.

So, how does this affect IoT architecture? Simply put, much of the core of IoT will happen at the edge. This may appear counter-intuitive to many who believe that data collection will be distributed and all processing will converge to centralized cloud locations. HPE believes differently, and Wikibon concurs:

“. . . there are seven reasons to not send all the data, back to the cloud. That doesn’t mean you can’t or you shouldn’t, it just means you don’t have to.” Dr Tom Bradicich, HPE

The seven factors guiding IoT architecture as identified by HPE, with Wikibon explication, are:

  1. Latency. Nothing travels faster than the speed of light, including data. That means that the time required to perform an action cannot be less than the time for a sensor signal to send and a control signal to return. A good rule-of-thumb for minimum round-trip latency is 1 microsecond for every 50 meters of fiber round trip. Put another way, for an autonomous car traveling at 60MPH, the car’s “brain” better not be more than about 50 meters from the car if you want to make speed and direction adjustments every few centimeters.
  2. Bandwidth. According to Eric Starkloff, the average OT customer has acquired over 22 exabytes of data. That’s equivalent to a million years of streaming HD video, way too much to presume that all data moves to a single location in the cloud. Bandwidth is not infinite. To maximize the productivity of available bandwidth, users increasingly will have to make concrete decisions regarding which date to move and when to move it.
  3. Costs. Whenever data is moved, processed, or stored, costs are incurred. To ensure practical strategies for IoT data management, users have to develop concrete approaches to valuing data. No common conventions for establishing data value exist, but it’s an era of intense focus in the tech community (including Wikibon). However, generally speaking, IoT costs will be moderated if work is performed at the edge.
  4. Duplication. Because of latency and bandwidth issues, significant IoT data will be duplicated at the edge, in centers, and in between. However, policies for record management, data retention, and deduplication won’t be superseded by IoT latency, bandwidth, and cost concerns. Instead, business policies related to data have to be rewritten to accommodate the realities of IoT data duplication, including data that is stored at the edge.
  5. Threats. Data in motion is data that’s vulnerable. The challenge, however, is that the physics-based latency, bandwidth, and cost constraints facing IoT applications don’t magically disappear when moving, processing, or storing security-related data. Edge computing won’t work if security-oriented processing is centralized. Wikibon believes that blockchain technologies are a promising direction for solving the distributed security needs of IoT.
  6. Corruption. Data in motion is not only vulnerable. It’s also subject to corruption in the form of data loss. For traditional applications based on application-specific, digital data structures, data corruption generally is manageable. But for IoT applications that translate analog into digital data and then move and process it, fidelity of sound, visual, environmental, and other data can be crucial to system performance. The best way to avoid data corruption is to simplify routes from source to destination, minimizing hopping and avoiding congestion, which is easiest if crucial, time-sensitive IoT data is handled at the edge.
  7. Geo-fencing. While not a physical limit, another reality is the impact of government dictates regarding privacy. Data can’t move faster than the speed of light, and in many nations, certain data can’t legally move across borders.

IT and OT Are Converging

Operational technology (OT) has been at the forefront of this effort for a few decades, but historically OT systems and networks have been largely segmented from IT systems and networks, for reasons like security, predictable real-time performance, and institutional control. However, as OT-technologies like sensors and actuators get smaller, cheaper, more easily networked, and considerably more functional, IT technologies like machine learning, blockchain security, and cloud-based development become more important to OT. To cross the IT and OT divides, vendors are beginning to enact strategic partnerships. Users, like Ford, General Electric, Mattel, are taking steps to ensure that institutional constraints don’t limit OT and IT integration as they seek to better serve their markets with digital assets.

“. . . big analog data . . . is the data that is pent up in things. And . . . National Instruments, can extract that data, digitize it, make it ones and zeroes, and put it into the IT world where <HPE> can compute it and gain these insights and actions.” Dr Tom Bradicich, HPE

Action Item: The data predictability, fidelity, and volume assumptions employed for architecting traditional IT systems will require a makeover for planning, building, and running IoT systems. HPE has identified seven factors which will shape IoT architectures, favoring the edge and multiple processing centers. Other factors may also impact IoT architectures, but start communicating and applying these factors today in IoT system plans and designs.

Book A Briefing

Fill out the form , and our team will be in touch shortly.
Skip to content