IoT: FROM COMPLEXITY TO SYMMETRY · Internet connected things. The Internet of Things (IoT) is no longer a buzzword but the inevitable state of the industry. At some point in time,

Cannot be reproduced without permission from Rumble1 /DATA AS A SUPER Power. ~ [email protected] ~ 800 926 0556

RUMBLENOW.COM

IoT: FROMCOMPLEXITYTOSYMMETRYA perspective on the complexities and obstacles

in the Internet of things and how to successfully

build scalable enterprise IoT solutions.

THE COMPLEXITIES OF IOT

You, your business, and your customers have heard and seen the proliferation of Internet connected things. The Internet of Things (IoT) is no longer a buzzword but the inevitable state of the industry. At some point in time, the economies of silicon, software engineering advances, and most importantly, the development of use cases would have come together to build an industry out of IoT.

The cost of silicon for example has followed the industry rules: Moore’s Law, Dennard Scaling, and Koomey’s Law. That is the number of transistors in a chip and the performance per watt of energy has continued to progress (albeit at slower rates). The world of IoT has been last to appreciate such technologies since most of the embedded world is based on computational devices using leg-acy 8-bit micro-controllers, PLDs, or PICs. We have reached a point now where silicon cost and area are efficient enough to allow for a significant and profound amount of computational power to be placed into these legacy chips. Now we have the ability to connect what was once unconnectable.

Advances in hardware are the basis of software proliferation. If modern society were to use a classic 16-bit computer architecture from the 1980’s, we wouldn’t have such fundamental features as a multi-tasking operating system, HD resolu-tion, wireless peripherals, or even the Internet. Today, a device with similar costs comes with significant capabilities: from sheer performance to hardware security and a plethora of IO options. This is what opens the world of IoT and claims of 20 billion to 1 trillion connected devices.

This is the first stop when we talk about the complexity in IoT. An engineering solution usually doesn’t start as complex nor do engineers and architects list complexity as something you want to maximize. Rather, a system becomes complex based on three factors:

1. Significant coupling and interfacing between modules, components, and devices.

2. The handling of unpredictable factors that involve the analog world and human or natural behaviors.

3. Changes in format and phase of data through disparate communication fabric, software interfaces, and network protocols.

All of these issues become significant for the IoT world. As we will see, mov-ing data from a simple sensor to a cloud data-lake comes with significant complexity that can’t be trivialized.

More so, the complexities of building a scalable, maintainable, and cost effective IoT solution can become more complex than cloud e-Commerce sites, autonomous vehicles, and even the most advanced machine learning algorithms running on state-of-the-art GPUs. The fact is the IoT world is a superset of all those examples.

Take for instance asset tracking. Suppose your company maintains a fleet of utility vehicles and within those vehicles you have different but critical assets. Perhaps this is for an electric utility company and those assets include com-puters, test equipment, radios, etc. The problem to solve is locating where assets are in real-time, alerting someone in the event an asset is missing, and allowing but recording which assets to move from vehicle to vehicle. The goal would be that the cost of such an IOT deployment would be made up for in recovering the loss of equipment and service downtime.

Perry Lea


RUMBLENOW.COM

THE ADVANCES IN SILICON DESIGN and process have led to significant functionality and performance in what are new classes of micro-controllers.

Twenty years ago, 80c51 micro-controllers were the main-stay of embedded compute and provided control and pro-grammability at the edge. Today, for nearly the same cost you can purchase a system on chip with 261 times the per-formance and 6000 times the transistor count.

An abundance of IO capabilities, and the ability to run out of the box Linux, allows these devices to bring connectivity to what was unconnectable 20 years ago.

1998 2018Product Intel 80C51 Microchip SAMA5D2

CPU Single 8-bit ALU @ 33MHz

32-bit ARM A5 @ 500 MHZ with NEON Media Processing FP

Memory 64K addressable 4 GB DDR3/LPDDR3

IO Full Duplex UART, 32 GPIO

1024x 768 LCD, 10/100 Ethernet, Two USB 2.0, SPI, CAN, UART, 2-wire, 12C, ADC, 128 GPIO

Operation Systems None Linux and modern RTOS’

Die Size About 3000 transistors in .33mm2

About 20,000,000 tran-sistors in .4mm2

Power .11 mW/MHz .09 mW/MHz

Performance About 3 MIPS 785 DMIPS

Cost Roughly $4 in 1998 $3.97 in 2018

The turgid complexity of such solutions manifest when we break down the compo-nents. Assets could be managed using simple battery-based beacons that com-municate to edge gateways in each vehicle. The gateways are responsible for long range WAN communication as well as governing the near range communication to these assets. The designer may choose a certain cloud provider to receive bea-con data and perhaps have some real-time dashboard to display them on a map. There may be additional rules that are built in the cloud to identify when assets go missing.

At a 50,000 foot view the architecture seems simple and it should be – from an end user point of view. However, the complexity stems from realizing the complexity of building a scalable solution. Consider:

Asset tracking:• What beacon technology should be used: iBeacon, Eddystone, proprietary

900 MHz? • What is the range that items should be tracked?• Are beacons actively polled or passively advertised?• How are devices provisioned?• Does the beacon need to be encrypted?• How do we optimize the battery life?• Do we consider a point to point or mesh topology at the PAN level?

Gateway Edge Processing:• How much edge computing is necessary?• How is the gateway secured?• How are firmware upgrades managed?• What WAN radio technology needs to be incorporated?• How is the gateway secured and environmentally protected?• Does the gateway need to perform rules engines or caching?• Is this a custom board or OTS?

WAN:• What long range communication strategy should be used?• What are the carrier and overage charges?• What is the coverage and range?• Should we consider SDWAN?• Does the signal work in all environments such as rain, fog, or within

parking structures?• What protocol is used for edge communication: MQTT, CoAP,

Web Sockets?• What is the effect of protocols on battery life?

Cloud:• Which cloud service or services do we use?• What is the backend protocol and scaling?• Do we consider a server-less design or containers on a PaaS?• Do we need long term data retention services?• Do we need a DevOps team to manage cloud deployments?• What are the charges as we scale to more vehicles?• Do we need to tie into a legacy on-prem asset database?• These questions (and many more) are all required to build a solution

that scales from a proof of concept with one vehicle to potentially thousands of vehicles spread geographically.


RUMBLENOW.COM

IoT architecture is not cut and dry. There are 100 MILLION COMBINATIONS of choices of how you will capture, move, store, and process data from the sensor to the cloud and back. We are at high-point in the IoT hype cycle. Many organizations are capitalizing on the hype and with a myriad of solutions and products. Finding the right combination of technologies to solve -the problem- is key in building IoT systems. HOW HARD CAN THIS BE? Often is the case, where rightfully intelligent people will under-scope what is unbe-knownst to them. IoT scaling is a very complex problem. A company with history and skills in building cloud components often overlooks the complexities and constraints of getting data from edge sensors to the cloud in the first place. This is because the IoT is built on many domains and sciences that require expertise not only within a specific silo, but also an understanding of how each domain affects one another. A typical cycle plaguing the IoT industry is that a company well-versed in their market such as routing, develops less than robust solutions for attaching routers to sensors or cloud components. Often these solutions are less than robust. There have been cases where experts in building expensive industrial sensors for factory automation use a single board development kit running an insecure Linux distribution to communicate to a cloud provider using a less than optimal protocol. Such designs lead to systems that typically can’t scale, defeat their OPEX goals, or become security liabilities. Consider how many choice points there are in building an IoT system. For example, there are over a dozen different near range communication technologies to get to the IoT edge. Some include Bluetooth, Zigbee, EnOcean, Dash, Zwave, proprietary 900 MHz, even LTE Cat M1 or NBIOT. Next you have a choice in WAN protocols from LoRaWAN to Sigfox to LTE. Add to that a mix of over 100 cloud providers. Within each cloud there are choices in WAN protocols, configuration of Spark and Streams, different ingestors like Kafka and Flume, along with a myriad of storage options. A customer will want to analyze and do something with the data that is collected. There are over 200 different analytic tools and packages as well as individual domains of AI. Choose wisely.

To add to the complexity, often industrial and brown field industries adopting IoT rely on legacy systems for CRM, ERP, field service management (FSM), inventory control, product lifecycles management, and supply chain management. Such industries also rely on legacy communications methodologies like MODBUS, 2-wire, and RS485. These types of mission critical systems are expected to be reused and capitalized versus replaced. Industrial IOT is built on legacy systems. Often you will find DB2 databases and proprietary OT systems. How do you combine, reuse, and build IoT systems that were never intended to be connected?

What are examples of unintended consequences in IoT and how do these domains of engineering need to work together? Here are some examples:

CASE STUDY A:Launch: A large US based shipping and package delivery service optimizes a system to connect their drop boxes to allow carriers to know when pickups are due. The system is optimized for data delivery to the cloud and uses advanced analytics to predict when deliveries and pickups are needed.

IOT SOLUTIONS AREMULTIDISCIPLINARY


Result: The system requires an always-connected infrastructure at the edge where power isn’t always available. Communication and cloud costs rise. Initial KPI to optimize carrier collection services is not met.

CASE STUDY B:Launch: An international manufacturer of solar arrays builds solutions with little attention to connectivity. Solar arrays need communication back to the cloud to rate power usage and credit extra power fed back to the grid. Manufacturer optimizes for cost and builds solution using development kits to provide connectivity.

Result: System cannot scale and is only working in certain European regions. System relies on WiFi connectivity through residential customer WiFi which is unreliable. Customers complain of uncredited power. USE CASESExamining selected real-world use cases reveals the details of IoT deploy-ments and where less than obvious intricacies can lead to failure. This section will examine agricultural, smart-lighting, healthcare, and multi-site manufacturing IoT. The take-away should be that each IoT solution is unique from the sensor to cloud. Each solution needs to optimize a different customer KPI.

MULTI-SITE MANUFACTURING Industrial 4.0 is all about the digitization of brownfield industry. Manufacturing has been automated since the first industrial revolution, but only today do we have the technology capable of tying every process, device, and material together on the facto-ry floor. The key performance indicators of a successful Industrial IoT solution should factor:

1. x1% reduction in OPEX (throughput, capacity)2. x2% improvement in systems uptime3. x3% reduction in material waste4. x4% returns improvement and warranty reduction

Industry is also built upon legacy systems (some machinery being over a half a century old). They also demand near 100% uptime of the equipment as well as the resourc-es and supplies needed to feed such equipment. This is the modern production line which is analogous to a pipeline. The radio diagram shown below illustrates the six vectors of IoT complexity.

RUMBLENOW.COM

1086420

Complexity ofAlgorithms

CommunicationsComplexity

Legacy IT Systems

Number of IoTNodes

Business Critical

Integration withPartner System

DATA AS A SUPER Power. ~ [email protected] ~ 800 926 0556

RUMBLENOW.COM

These systems are legacy SCADA infrastructure elements such as on-prem networks, Microsoft Access, and even DB2 databases. Often these systems have been used for years and have near 100% uptime. Typical machinery will rely on 1980 or 1990 factory automation. This includes legacy but popular near range communication us-ing physical wiring such as RS485, Modbus, and 2-wire. Furthermore, OT staff will be reluctant to retire this infrastructure as the return on capitalization costs haven’t matured. In building IoT solutions surrounding these systems, the design must com-plement them and not replace them. Industrial systems usually have a refresh cycle of 10 to 25 years. Along with these constraints, an IoT solution must also manage multi-site and multi-en-tity roles. Take, for example, a global manufacturing company building automotive parts. The company may have different factories spread across the US, Mexico, Europe, and Asia all using different equipment and building different products. Components may be built at one location, shipped to final assembly centers worldwide and then stored in distribution centers in other locales. Such geographic spread makes Just-In-Time man-ufacturing complex and poorly responsive to unpredictable changes in supplies, equip-ment downtime, or even weather. Their suppliers, as well as customers, also want a prior or near real time knowledge of supply or distribution issues. This company needs to optimize across all KPIs which means improving quality and production through IoT. These constraints and complexities can be managed and re-turn positive results to the business using prescriptive design and architecture, as we will talk about later.

HEALTHCAREHealthcare represents a significant opportunity for IoT to add value, safety and patient benefit. Examples of healthcare deployments include:

• Remote patient monitoring and in-home care using healthcare sen-sors and devices.

• Patient tracking and location devices.• Drug and equipment location tracking and monitoring.• EMT and mobile health solutions in remote areas.• In these cases, the KPIs revolve around the patient. The patient is the

basis of all surrounding metrics:

1. x1% decrease in overall health costs.2. x2% decrease in patient response time. 3. x3% increase in patient satisfaction and recovery.

Healthcare systems are heavily regulated with HIPAA and legal require-ments for patient safety and confidentiality. Devices directly monitoring or attached to humans will also have requirements governed by federal agencies like the FDA. Unlike industrial IoT, healthcare systems are typi-cally more amenable to newer or replacement systems. IT and legacy sys-tems will also be easier to interface with versus brownfield manufacturing. Electronic health record (EHR) systems follow standards and HIPAA regu-lations. While modern, EHR systems are not used everywhere in medical care, certain incentives exist to promote adoption. Additionally, differences exist in healthcare standards globally. Some of the additional complexities of healthcare IoT are in communication patterns. For example, remote patient monitoring requires some connec-tion between the patient and Internet. Often is the case where the patient may not have adequate or any WAN communication. A patient may also incorrectly use the device, or the near-range communication medium is poor. The results are less than satisfactory experience and the cost may negatively impact the bottom line.

The Six Vectors of IoT ComplexityStudy several IoT use cases and normalize nebulous complexity factors into separable vectors that can be compared.

Legacy IT Systems The normalized amount, quantity, or number of different IT systems based on 10+ years technical maturity. Exam-ples include DB2 databases, and legacy on-prem IT equipment.

Number of IoT Nodes The number of IoT devices or endpoints that must be managed, measured, and governed.

Business Critical The significance of downtime of a system managed by IoT to the overall business.

Integration with Partner System

The normalized instances and complexity of partner and customer IT systems such as various clouds sources, external databases, and unmanaged interfaces.

Communication Complexity

How easy it is to reach an IoT endpoint. This includes the complexity of proprietary and legacy physical communi-cation such as Modbus, 2-wire, proprietary 900MHz, and RS485.

Complexity of Algorithms

This includes the overall complexity and size of software solutions and data. This may include multiple edge, fog, and cloud instances, machine learning and training, computer vision, data analytics, multi-OS, multi-architecture, and deeply embedded systems.

1086420



Legacy IT Systems

Number of IoTNodes

Business Critical

Integration with Partner System


RUMBLENOW.COM

AGRICULTUREAgricultural IoT touches every aspect of food production. From seed and soil treat-ment, to irrigation and resource optimization, to machinery preventative mainte-nance, to livestock health and location, to the logistics of food transport every aspect of agriculture can be optimized with IoT deployments.

Agricultural IoT solutions share some components of many other IoT solutions but are also quite different. For example, the key performance indicators in agricultural IoT include:

1. x1% decrease in OPEX2. x2% improvement in yield 3. x3% decrease in equipment downtime4. x4% improvement in food safety

Agriculture, at a fundamental level, suffers from the effects of environment predict-ability and repeatability. That is, plants, animals, and nature itself are greatly unpre-dictable. Weather for example, is tough to predict at a micro-climate level (down to the quarter mile). Animal and plant health also vary wildly. Some animals or plants may not all have similar growth rates or harvest schedules. Animals get sick, and they need to be identified early before an entire head of cattle are affected. Farmers also need to optimize feed and pesticide (and in many cases not use pesticide at all) as these affect the bottom line.

At a technical level, communication with agricultural IoT can be challenging. While widespread LTE may blanket the US, not all areas are covered, and these are some of the most fertile and widely used agricultural areas. What WAN option exits? Is there a uniform WAN option to blanket the US. Does the integrator need to consider alternatives such as satellite or a caching model? Consider the complications of tracking a herd of livestock. There may be thousands of head of cattle on a feed lot or thousands of free ranging animals in the foothills. How will they be tracked effectively? Does the system integrator consider a standard mesh fabric or a proprietary 900MHz medium range LAN?

Consider issues with powering agricultural IoT. What are the options of en-ergy harvesting at the edge? What are the effects of data communication on battery life? For example, consider using injectable sensors in poultry to mon-itor health, weight, movement, and location. Such sensors need to be at a cost point of being nearly disposable but also have no contaminating effects from electronics of batteries. If the intent is to report on poultry health to isolate disease as soon as possible how will a continuous stream of data from every trip to the cloud affect the device power? Or is there a better architecture?

SMART LIGHTING / SMART BUILDINGOur final use case is smart lighting and connected buildings. This case can extend to other smart city and smart infrastructure deployments. In this case, we consider connecting what had at one time been the un-connectable world of building controls and automation.

Smart lighting and smart construction of any kind suffers from their own in-dustry constraints and are also measured on unique KPIs:

1. x1% reduction in energy costs for property owner.2. x2% improvement in response time to failing or would be failing building

infrastructure.3. x3% improvement in system uptime.4. x4% simplification of install or retrofit.

While saving costs and improving uptime are similar to other use cases, building infrastructure and lighting is a unique market where the IoT system manufacturer may not be the system installer. Rather, these industries rely on various contractors for installation and building service providers for main-tenance and upkeep. For example, a building service provider may require various dashboards and preventative maintenance solutions for several years. In contrast, the electrical subcontractor requires fast, simple, and methodical installation with near-zero IT knowledge involvement to secure and provision a new smart light.

Another factor is communication standards. Currently there are widely vary-ing standards and many proprietary communication protocols for smart light-ing. These include sub gigahertz systems, various 802.15.4 standards and Z-Wave. For the same reason that it is easy to change a lightbulb – it is very difficult to change a smart lighting bulb. The device needs to be of the same communication protocol as its neighbors and control system and it needs a complicated provisioning process. The industry is striving to get to a standard approach that is both easy for a contractor to install and provision but also secure.

1086420



Legacy IT Systems

Number of IoTNodes

Business Critical



RUMBLENOW.COM

FROM SENSOR TO THE CLOUDThe following is a dataflow example of moving bytes of data from a temperature sensor on a remote edge node to the cloud. Along the way, we explore the various transforms, data manipulations, forks-joins and splits of the data. The intention is to illustrate that professional IoT deployments are complex and can’t be trivialized. Let’s begin with a simply thermocouple sensor measuring some fluid heat dissipation.

We start with an analog change in environmental conditions that cause a change in voltage across two metals in the sensor.

1. This change can be measured by some low-cost amplification and ADC circuitry. The sample is measured at some periodic interval and captured in a small embed-ded SRAM (and buffered). The value is literally a 2-byte quantity.

2. The information is then transmitted from the sensor as an advertised beacon to save energy. A gateway using a Bluetooth 4.2 BLE PAN will receive the data. The data is encrypted and marshalled through 5 layers of hardware and software stacks and packetized into a 2.4 GHz signal that is modulated and hops frequen-cies as the environment is noisy.

3. A gateway device hosts the receiving BLE componentry and unrolls the layers of BLE frames and packet structure to uncover the 2 bytes of data.

4. The gateway then wraps the two bytes in an IP framed packet and then overlaps the packet with an MQTT structure. The data is then re-encrypted using a TLS 1.2 protocol.

5. The data is then passed to buffers to an LTE modem which is awakened with a new message. The message content is wrapped into a E-UTRAN protocol stack. The modem has established a bearer relation with a E-UTRAN tower and carrier. Data is multiplexed and sent to the awaiting E-UTRAN.

6. EPC hardware managed by a carrier now is marshalled to the Internet backbone.7. Data manages its way to a provisioned cloud provider. An ingestor service in the

cloud receives the data and verifies the identity and access privilege of the mes-sage.

8. Upon verification, the data is decrypted using TLS 1.2 and manages its way to an MQTT topics branch and forked to a real time stream analyzer: Spark.

9. Data is then analyzed using a time-series correlation. Interesting data is stored in glacier storage while less than interesting data rotates in a ring buffer in a da-talake.

10. Data is then presented to a user in an OT dashboard. 11. Along the way many things can go wrong. Information and communication can

be lost or incorrect. The complexities grow when data must be split and forked to other services along the way. Data sources can also come from multiple end-points and travel to multiple clouds and use services and data from multiple da-tabases. When we scale the solution to many nodes and multiple regions the complexity increases further.

The IoT world is more complicated than developing a single embedded system or a cloud application. It must consider all those elements and more.

IOT NEEDS MULTI-DOMAIN EXPERTISE.


RUMBLENOW.COM

STANDARDS TO REDUCE COMPLEXITYIndustrial and community standards are generally perceived favorably in the technical community and appreciated by consumers and customers. Standards, whether contrived by industry or mandated by legislature, attempt to drive to uni-formity for various reasons:

1. Safety: FMVSS for US automotive safety standards. Dictates such things as seatbelt locations and tensions to windshield wipers.

2. Interoperability: ASME standards for pipe fitting.3. Longevity: NEMA standards on lighting. Incandescent lightbulbs from the

1920’s can still be used in today’s lamps.

The explosion of the Internet of Things buzzword has created a proliferation of technologies from APIs, to protocols, to wireless communications. This affects customer value in safety, interoperability, and longevity. A successful standard doesn’t impede innovation or stifle productivity, rather the opposite is true. Often is the case that standards allow for a proliferation of technology and opening up of new markets.

Here we look at some options in the IoT to build on and reinforce standards.

NEAR RANGE COMMUNICATION (PAN)BLUETOOTH SIG (WWW.BLUETOOTH.ORG) Membership: 20,000Bluetooth SIG was formed in 1998 and provides the standards for certification of Bluetooth near-range communication. With Bluetooth 5 and the new mesh stan-dards, it is poised to be the most robust and flexible near range communication system available. It has strong presence in sensors, peripherals, consumer elec-tronics, and was potential to be the defacto lighting mesh standard.THREAD (WWW.THREADGROUP.ORG) Membership: 182The Thread Group defines the standards around the 6LoWPAN protocol based on 802.15.4. This protocol is an IPV6 packetized communication network for near range communication allowing every IoT device to have a unique IP address. ZIGBEE ALLIANCE (WWW.ZIGBEE.ORG) Membership: 446This working body governs the Zigbee protocol based on 802.15.4 standards but is not IP based. This standard is flexible and allows for mesh networking prevalent in home and consumer electronics. MODBUS ORGANIZATION (WWW.MODBUS.ORG)Membership: 105Modbus is a legacy brownfield wire-based interconnect prevalent in industry and manufacturing since 1979. The Modbus organization has been established gov-erning the standard wince 2004 with new interfaces like Modbus over TCP/IP and UDP.

LONG RANGE COMMUNICATION (WAN)WEIGHTLESS (WWW.WEIGHTLESS.ORG)Membership: UnknownWeightless is a long-range communication LPWAN protocol governed by this special interest group. Weightless is heavily used in IoT healthcare, transporta-tion, vehicle tracking, and industrial environments. LORA ALLIANCE (WWW.LORA-ALLIANCE.ORG)Membership: 419Founded in 2014, the LoRa Alliance is a non-profit consortium defining the stan-dards for its sub-gigahertz long range communication systems. Typically used in M2M and smart-city deployments, LoRa has gained traction in smart agriculture and energy IoT. 3GPP (WWW.3GPP.ORG) Membership: 5723GPP is a collaboration between groups of telecom standards bodies. Nearly every carrier, equipment manufacturer, and chipset designer are in 3GPP. This multi-regional and multi-national body defines the standards around cellular long-range communication for voice and data traffic. They are the body developing the release model of cellular technology such as GSM, CDMA, and LTE. They are also deeply involved in 5G standards. SOFTWARE STANDARDS AND PROTOCOLSOASIS (WWW.OASIS-OPEN.ORG)Membership: 300OASIS is the standards body formed in 1993 that governs IoT and device commu-nication protocols such as AMQP and MQTT. Originally designed and used by IBM, the standards are now open but specified through OASIS. Because of such wide-spread use this standards body is significantly important to the IoT industry.OMG (WWW.OMG.ORG)Membership: 250OMG standards for Object Management Group which was established in 1989. The original group involving Apple, HP, IBM, and Data General are best known for their work in UML and CORBA software standards. They have also worked on standards around software defined networking (SDN) and IoT data management.IPSO (WWW.IPSO-ALLIANCE.ORG)Membership: 60This organization was formed in 2008 with the goal of putting IP addressability on each and every connected object. The group promotes standards of interopera-bility with IP as the foundation. This group complements the Internet Engineering Task Force.INTERNET ENGINEERING TASK FORCE (WWW.IETF.ORG)Membership: 1200The IETF is a leading and important standards body governing all aspects of TCP/IP and protocol stacks including: 6lo, LPWAN, and IPV6. It has existed since 1986 and includes the top corporate, government and academic membership setting the defacto standards of network communication.

RUMBLENOW.COM

EDGE PROCESSINGOPENFOG (WWW.OPENFOGCONSORTIUM.ORG)Membership: 55OpenFog is a relatively new non-profit organization trying to build standards on the burgeoning world of fog computing. To date several architectural reference specifications have been drafted and membership is growing with sponsors such as Dell and Cisco. The architecture contains specifications for security, containerization, deployment, and load shifting/balancing.EDGE X FOUNDRY (WWW.EDGEXFOUNDRY.ORG)Membership: 50Another new organization building on standards around edge compute. Orga-nized through the Linux Foundation, the organization is responsible for defining standards around edge micro-services, middleware, rules engines, alerting, logging, and registration.

UMBRELLA ORGANIZATIONSINDUSTRIAL IOT CONSORTIUM (WWW.IICONSORTIUM.ORG)Membership: 258The IIC was formed in 2014 by AT&T, Cisco, GE, IBM, and other companies involved in industrial and manufacturing IoT. The group is not a standards body but deliver reference architectures for manufacturing, healthcare, transporta-tion, and smart city IoT.

IEEE IOT SIG (IOT.IEEE.ORG)Membership: IEEE 100,000’sThe IEEE is the standard of standards organizations. They are responsible for the mindshare of nearly every technology. This multi-disciplinary SIG within IEEE influences protocols like 802.15 and 802.11. They sponsor the leading in-dustry forums and conferences on IoT including the World Forum-IOT.

PROFESSIONAL SOLUTIONS Even with standards bodies and blanket organizations, deploying successful IoT solutions that scale, that are secure, and that deliver on KPIs is a non-trivial prob-lem. We have learned that IoT development requires the expertise of multiple engineering domains from power electronics and sensor physics to cloud XaaS DevOps management. A company chartered with capturing or extracting value from their operations through IoT should ask themselves: “how deep do we want to go?”. If the answer is to deploy at scale at enterprise levels, then it makes sense to use professional services that are experts in IoT and edge computing to offset companies that are truly experts at their product.

As professional in IoT development, we find that no two IoT deployments are the same. We have learned that there are over 2 million choice points in architecting a solution and each solution has different constraints. If we look at the four indus-tries we evaluated, it’s clear that the requirements and constraints shown in the radio diagrams are unique. Certainly, portions of IoT solutions may be re-used, but holistically we find systems are rarely identical.

Industry needs to understand that there is no “silver bullet” to build up an IoT solu-tion. There is only hardware and dedication from expert bodies that understand the cross-domain effects of enterprise level IoT.

RUMBLE helps companies overcome the complexities of IoT implementations le-veraging and leading teams of interdisciplinary experts to design, build, and imple-ment custom IoT solutions that deliver on business objectives. Leveraging an IoT Strategy is more than connecting devices, it is a shift in thinking about the value proposition of your business. RUMBLE’s experts can help you re-envision your business based upon the opportunities presented by IoT. Whether your industry is manufacturing, healthcare, transportation, agriculture, oil and gas, or retail, we have the capability to help you improve productivity and drive revenue with IoT. We serve as your IoT quarterback managing the various players resulting in a suc-cessful ROI. Let’s start exploring where IoT can take your business.

GARTNER PREDICTS THAT 50% OF ALL IOT PROJECTS WILL REQUIRE THIRD-PARTY INTEGRATORS AND TOOLS.

10

8

6

4

2

0



Legacy IT Systems

Number of IoTNodes

Business Critical


Multi-site Manufacturing Agricultural Healthcare Smart Lighting


WRAP UP It is important to realize that our industry it built on three laws: Moore’s Law, Koomey’s Law, and Dennard Scaling. These trends in hardware have enabled the IT industry to take what would have been an extremely limited micro-con-troller that could only communicate through a UART twenty years ago, to now running multiple real-time operating systems, fog and edge Lambda, run con-tainer or VM instances, and use every modern programming paradigm for the same cost. This has allowed a proliferation of software capabilities to run at the edge and connect to the Internet hence the Internet of Things.

With this ability comes complexity and the industries best served by IoT also have a prolonged IT legacy of their own. A one stop shop cannot solve all IoT problems. Each solution is industry specific and constrained by different as-pects of the business. There is no silver bullet, cloud XaaS system, middleware, gateway, or sensor that will satisfy all industries.

We as industry practitioners must strive for the adoption of standards and in-dustry best practices. With standards comes the opportunity for uniformity and democratization of infrastructure. Additionally, a IoT adopter must understand where to assign professional services versus an in-house build. The complexity of IoT requires multiple domains of expertise that can’t be trivialized. Poorly in-vesting in one component of the IoT chain creates solutions that can’t scale, are not secure, and break down with slight variance in an otherwise analog world.

RUMBLENOW.COM


ABOUT RUMBLE

RUMBLE, is dedicated to helping clients harness the power of IoT. We lever-age data to support near real-time decision making and action. The result is performance improvement and revenue growth.

Rumble was formed because IoT Architecture and Implementation are complex and a specialty that few companies possess, Rumble being one of them.

Progressive businesses, and businesses that will outperform their compet-itors, are connecting real-time operational performance to the people, or machines, that can act to improve outcomes. Agile and innovative business management requires access to data. Most companies have siloed data, stagnant data, reporting lags, and disconnections that hinder impactful de-cision making and action. Rumble specializes in tools, like IoT solutions and situational intelligence software, to break down these barriers and present the data in a near real-time, actionable format. RUMBLE is also a Premier Partner of Coolfire Solutions, a leader in situational intelligence software.

We have the expertise, the experience, and the tools to deliver the promise of IoT and performance improvement to your organization. Let’s get started with a conversation.

[email protected]

ABOUT THE AUTHOR

Perry Lea is a 27-year veteran of the technology industry and strategic partner of Rumble. He served as Distinguished Technologist and Chief Ar-chitect of the Imaging and Printing Group of Hewlett Packard. There he architected and technically steered the design of over 60 product lines in a $20B industry. He also worked with HP Labs and various industries in a processor technology, security and memristor design. He then served as Distinguished Member of Technical Staff and Director of Strategy at Micron where he developed the world’s first non-Von Neumann computing tech-nology to directly address the inefficiencies of computer vision, machine learning, and big data analytics. Later he worked as Director of Technology for Cradlepoint building a successful IoT business and technology line. He now serves as CEO and Founder of Computational Vision providing pro-fessional services and consultation in areas of emerging compute, IoT and edge computing, and business development.

He is author of over 50 patents and has recently published “The Internet of Things for Architects” through Packet Publishing. He holds engineering degrees in computer science, computer engineering and a post-graduation degree in electrical engineering from Columbia University.

[email protected]

RUMBLENOW.COM

Documents

IoT: FROM COMPLEXITY TO SYMMETRY · Internet connected things. The Internet of Things (IoT) is no longer a buzzword but the inevitable state of the industry. At some point in time,