18 August 2017

Whoever is responsible for the network, they will have to ensure that each and every IoT device – be it a smart meter, connected vehicle, security camera, etc. – can connect to the network when, where and how often users expect it to.

You have probably heard all the headline-grabbing forecasts of the billions of devices that will be connected in the Internet of Things, and the trillions of dollars that the global market is expected to bring in for vendors, suppliers, network operators, etc. And you’ll probably also be aware of the wide range of applications enabled by the IoT, from utility companies deploying smart meters and snacks firms monitoring their nationwide vending machines, to train operators managing their rolling stock and healthcare services keeping a remote eye on their patients. But in all instances, somebody has to take charge of running the network. So whose responsibility is it?

“In many ways, an IoT deployment is pretty much like the internet: it’s a network of networks,” says Chris Mason, EMEA sales director and VP of business development at US-based wireless network specialist Rajant. “What tends to happen is that network providers, or operators, or the owners of devices have a boundary of responsibility for the communication devices. If you look at the IoT like a Venn diagram, it is really a combination of individual owners making up the whole.”

Telensa, the UK firm that provides wireless smart city control systems, is likely to agree here. Its VP of marketing Keith Day says: “There are a broad range of different networks covered under the term ‘IoT’, and the ownership and responsibilities for these paints a pretty diverse picture. Telensa’s focus is on smart cities. Here, responsibility is split between the city or municipality operations department, their contractors and the owners of the infrastructure.”

Earlier this year, Day published a blog that highlighted the rise of the “chief data officer”. He wrote that the idea of a smart city is based on the collection and analysis of Big Data, and if municipal authorities are to embrace the idea of “data-driven government”, they need to appoint CDOs to navigate their path to becoming a successful smart city.

While the idea of a CDO may be taking-off in US cities such as Chicago and New York which are said to have pioneered the role, in the UK there seems to be stricter demarcation around job titles such as ‘data scientist’, ‘data engineer’, ‘data administrator’, as Ed Thewlis of the DataShed suggests (see p3, What’s in a name?).

Other industry experts point out that for those companies who use cellular networks to connect their devices, the responsibility for managing the network is the mobile operator they have chosen. For instance, Greg Aston, product director at ETI Software Solutions, says: “Most of our IoT customers are currently using GSM-based networks and expect them to ‘just work’ in the same way you would for your mobile phone service.”

Whoever assumes responsibility for the network, one thing is clear: connectivity is king. As Theresa Bui, Cisco’s director of IoT strategy, points out: “It doesn’t matter how good the hardware of the device is, or how sophisticated the application or firmware that sits on it is – without connectivity, nothing is happening with that device. Connectivity is what turns a ‘thing’ into something that offers a continual service.”

US-based IoT platform provider Essence says all IoT network applications are just part of a larger system. As a result, the ultimate responsibility for their integration and ensuring that they run smoothly falls within the remit of an organisation’s IT functions. But Rafi Zauer, the firm’s head of marketing, adds that many, if not most, services are based on a cloud-based model. “In these cases, service providers (the customer) have an SLA with the technology vendor who are committed to 100 per cent uptime with 24/7 support. So it falls to the vendor to ensure that their systems are always working.”

Rittal highlights the first of many challenges here. Barry Maidment, the company’s south division IT sales manager, warns that as the IoT gathers pace, millions of machines, gadgets and household items are sending small packets of data to the cloud every second of the day. “The speed at which the data travels is dictated by the bandwidth and it is increasingly becoming overloaded,” he says. “Anyone whose business is totally dependent on the cloud needs full connectivity without internet interruption or downtime.”

Making the net work

Cisco’s Bui has already mentioned the importance of connectivity, and adds that an IoT network has to be able to provision, allow and monitor a connected device’s behaviour. If it can’t do that, it’s basically useless.

“When you are responsible for the network, you have to ensure that each and every device can connect to the network when, where and how often the users expect it to,” she says. “For example, a smart gas meter may only be expected to connect once every 24 hours, from a single location, and pass 19 kilobytes of data. But a car may be expected to connect to a network several hours per day, across different geographic zones (even countries), and pass hundreds of megabytes per day.”

Bui goes on to say that when it comes to connecting their assets for the first time, those responsible for the IoT should watch out for not being able to map connectivity to the device’s lifecycle. She says this could result in paying for needless usage prior to a go-live at a customer site, as well as unreliable connectivity at first use at the site. To help overcome this, Bui recommends device makers to enable some connectivity capabilities during manufacture so that the unit can be tested. There should then be no connectivity during shipping and distribution, but once the device arrives at its destination, it should have the ability to connect to the right network and start working immediately.

It is also crucial for users to establish what constitutes an IoT device’s ‘normal’ network behaviour. Without this, companies could end up with potential security risks and higher operational costs. “Establish the baseline for what is standard network behaviour for your devices so that you can be alerted from deviations from the norm to prevent cost overages and mitigate against potential security risks,” advises Bui. More on all that later.

For Rajant, Mason says enterprise users should begin their process of using the IoT by asking themselves what they want to achieve with all the data that will be gathered. He then emphasises that the technology deployed will always be a hybrid solution.

“It’s a combination of components – it may be components of our technology, someone else’s technology (such as an enterprise Wi-Fi provider), and a supplier of some low powered devices. So you have the devices, sensors or something generating a piece of data that needs to get a short distance from the critical environment in which it works to the point of collection. 

“Then the data is aggregated, and transmitted to reach the point of analysis which may be far away or local. We will just work to make sure that there is no barrier to the transmission of the data, and that its actual value is clearly understood by all parties, because that also determines other aspects that you put in, such as resilience, security, latency and multiple paths out. The end application always determines the components that you will need.”

Steve Bailey, MD at infrastructure and technical services specialist AIT Partnership Group, flags another potential problem with IoT projects: “Organisations are usually structured in silos with a facilities team and an IT team. Running an IoT network may require knowledge of protocols not familiar to IT teams such as BACnet and Modbus if legacy sensors, meters and assets are to be connected. Knowledge of M&E plant and where to place sensors and meters is usually a facilities role.”

As a result, Bailey says an IoT project will require close cooperation between the two teams, and that both may have to allocate budgets which can be a problem in some organisations. “It can be expensive to roll out an IoT project. You need a good ROI – ‘nice to have’ and woolly ideas don’t get off the ground.”

Telensa’s Day also expresses concerns about costs. He says that while the technology for smart cities and private IoT networks may be very similar, the way the funding is structured is “wildly” different. 

“Enterprise networks benefit from generous funding from the organisations that build and host them, and cellular networks benefit from individual subscriptions. But IoT networks have to operate at very low margins because each connected device tends to only make a very small cost saving or revenue contribution. 

“So you need to make the economics stack up. The challenge is how do you design an IT infrastructure, both network and applications, that will fit a highly constrained business case. How do you take cost out of the arrangement?

“The sheer volume of devices means that things have to be automated. And you need to take people out of the equation in terms of managing these things – both from a purely practical basis, but also from a cost perspective.”

Beware deviant devices

Automation is clearly a crucial aspect of the IoT. ETI’s Aston reckons that there is an expectation amongst users that everything will be plug and play: “In the consumer market, every new gadget that you purchase tends to require less user intervention to get online and functioning, so this experience is expected with a wide range of device types. This leads to the challenge of ensuring your infrastructure is as tightly integrated as possible and making sure you take advantage of automation wherever possible.”

Cisco agrees here when it says companies need to be able to monitor and control their devices in real-time. Bui says: “Since no company has the resources to manually do that, they need to be able to automate most of the connectivity management of their IoT devices. Look for a solution that enables you to simply type in your device’s standard behaviour.”

As examples of this “standard behaviour”, she says the IoT devices users deploy in a particular country should only work in that country, connecting once per day and with each session lasting no more than two minutes. They should pass about 2MB of data per month. “If any of your devices deviate from that pattern, disconnect them off the network and send a security notification to the support team. If the devices connect more than ten times per day, send an SMS to the operations team lead. And if they are passing more than 10MB per day, disconnect from the network and send an email to the operations team.

“Automating the connectivity management during a device’s lifecycle will ensure companies can scale their deployments quickly, ensuring reliability, minimising security risks, and all while lowering operational costs.”

Day lends his weight to the argument by saying that in the IoT space, wireless networks must be designed to operate over wide areas and run over decades with little intervention. “It’s also critical to have a lot of resilience built in, all while still being delivered and maintained at a low cost.”

When it comes to resilience, Rittal reckons one of the solutions that could be key is ‘fog’ computing. Maidment says that while cloud computing offers processing power and storage capabilities that few organisations can match and for a relatively low cost, bandwidth remains an issue. He explains that principle of fog is quite simple – it describes temporary storage and processing capability closer to home/the application. 

“Essentially, fog computing is a relatively small computer that gathers, caches and feeds data into and out of the giant servers in the cloud. The term ‘fog’ was apparently coined because it describes something closer to the ground and nearer the application. The ‘cloud’ by contrast is a distant facility.

“Fog computing is inclined towards an element of local processing and storage. What’s more, it builds greater resilience into the system and gives greater protection to the most sensitive data.”

According to Maidment, data processing is faster when a business has local storage facilities near the IT application. “When time is available, it then passes the consolidated data to the main cloud. The great advantage of fog, particularly for business critical functions, is that if your connection to the internet goes down and contact is lost with the cloud, you can still function as a business. It builds resilience into the system. And as more and more data is collected, analysed and processed through IoT, so the need for fog computing will increase.”

According to Aston, ETI can help overcome such challenges because of its Beamfly service and the software it provides in the IoT market. He claims the company has a “long history” of integrating systems to its core product, and ensuring there is a single touch point for users that can trigger coordinated interaction with a wide range of systems, such as billing, network provisioning, device management, etc.

But AIT’s Bailey advises against going for a proprietary software solution. Instead, he says users should choose one that can collect from the widest possible number of protocols. “Think about the future and consider building-in control features of IoT devices as well as data collection. MQTT is becoming ubiquitous but you may also need to connect to EnOcean, Modbus, LONworks, BACnet, SNMP, etc.”

Standards and protocols

There are several IoT technologies for connecting devices over long distances without internet connections. 

For instance in the UK, Arqiva is using Sigfox’s platform for its IoT network that was claimed to be the first in the country when it was first announced three years ago (see News, December 2014). 

LoRaWAN is another system and is being promoted by the LoRA Alliance. It is an LPWAN (low power wide area network) specification intended for wireless battery operated ‘things’ in a regional, national or global network. Networking+ did invite the alliance to participate in this feature but no reply was received.

Weightless is yet another M2M/IoT technology standard that has been around for a while now. Developed by a special interest group (SIG) whose board members include Accenture, ARM, Sony Europe and Ubiik, the latest version of the standard was officially launched in early August 2017. 

Weightless-P is an open standard for a high capacity LPWAN that is designed for performance. According to the SIG, when IoT connectivity technologies are being considered, users factor in parameters such as cost, battery life and range. It says while this is not wrong, it’s easy to overlook the importance of network capacity. “Capacity is not just about the number of simultaneously connected nodes,” states the group. “It is about mean data packet length, transmission time, frequency of transmissions and interference mitigation.”

Alan Woolhouse, chair of the Weightless SIG, says Weightless-P is narrowband modulation scheme technology while LoRaWAN is a spread spectrum technology. “What that means is a LoRa data packet occupies the entire bandwidth available in ISM [industrial, scientific and medical] spectrum, whereas a narrowband scheme communicates in narrow channels. You have a much higher spectral efficiency with a narrowband technology solution than you do with a spread spectrum solution and that translates into higher network capacity. Very roughly – depending on different regulations and implementations – Weightless-P has 100x the capacity of LoRaWAN (that’s based on a typical data packet size and a typical interval between transmissions).”

According to Woolhouse, Weightless-P “listens before it talks” and that enables it to schedule uplink traffic from end device to gateway. “If you can do that, you can guarantee that that piece of data will join the traffic stream in a way that avoids data collision,” he says.

While a steady stream of LPWAN technologies has come to the market, Bailey reckons the greatest coverage and bandwidth is still cellular. He says that demand for mobility has made Wi-Fi part of the corporate network but adds that for IoT, Wi-Fi deployment costs can be “prohibitive”. Nonetheless, as energy harvesting technology develops, he believes Wi-Fi may become the answer, especially if IoT is seen as part of the corporate landscape for IT projects.

But for Essence’s Zauer, the existence of varying competing standards for local connectivity often blocks the ability to offer an effective end-to-end solution in the residential IoT sector. “There remains a plethora of competing standards that create a barrier for service providers to give their customers an effective and flexible smart home experience,” he says. “It is important to maximise the benefits of each technology and utilise them where a particular network adds value. Relying on alliances and interest groups to create full interoperability can prevent the market from moving forward.”

Rajant seems to have a simple solution to any confusion created by all the various IoT standards and protocols.

“We are completely agnostic in terms of the kind of low-power end user devices that capture and disseminate information on a local scale,” says Mason. “From a very basic technology standpoint, Rajant’s approach has been to make sure that all of its devices act like a Layer 2 switch. Therefore it doesn’t matter what standards are used – as long as there is an Ethernet presentation coming out of the device, we will transmit the data.”

However, he does acknowledge that all the emerging standards and protocols do not make things easy for end users considering an IoT deployment. “When I look at organisations such as the Weightless SIG or LoRA Alliance, I see that both are relying upon the concept of open standards capturing the input of more than just their own developers, and emerging as a strong protocol as a result. I think there will probably be a winner, just in terms of the sheer number of networks deployed using one or two or another of these standards. But I also bet that if they are both strong runners, there will be somebody who solves the issue of getting them to work together.”

Day is likely to welcome that when he says: “Because the range of use cases and applications is so vast, a single standard could actually inhibit innovation and growth. For example, cellular and Wi-Fi technologies are based on completely different standards, but this has done nothing to limit their development and usage.”

Woolhouse adds to this by saying that a few years ago Sigfox seemed destined to be the winner for LPWAN technology and now LoRaWAN is in the ascendancy. “But today, people are generally fairly confident that NarrowBand-IoT is going to be a core technology that will prevail over the longer term. Weightless agrees with this, but we also feel that there is space for complementary, private network technology in unlicensed spectrum. 

“The actual mobile network operator business model is that they have a much higher ARPU [average revenue per user] from people using their mobile phones to access YouTube videos than they do from a machine that might, once in a while, send a few bytes. So it makes sense to offload some of that traffic to a different network in the same way that 4G and Wi-Fi are complementary today. We therefore feel that Weightless and NB-IoT will be complementary.”

Woolhouse says Weightless and NB-IoT have exactly the same benefits, and also points out that there is “no surprise” that Ubiik (the primary vendor of Weightless-P technology) and the 3GPP (3rd Generation Partnership Project) community have each chosen 12.5kHz as the channel width for their transmissions, as this offers optimum performance.

Furthermore, he emphasises that Weightless is a genuinely open standard: “The technology is available for developers to use on a FRAND [fair, reasonable, and non-discriminatory] and zero royalty free basis. That cannot be said for technologies like LoRa which use SEMTECH chipsets and chirp spread spectrum modulation schemes that are proprietary.”

Fit and forget? 

Once the IoT is up and running, what does the team responsible for it need to do when it comes to its routine operation? Security is of course always at the top of the agenda in all things IT, and perhaps even more so when it comes to the IoT. 

From a connectivity management perspective, Bui says the network team should ensure that proper monitoring gates are put in place. These should have the ability to disconnect devices from the network and alert administrators if unauthorised communication is detected. She adds: “If the devices have access to open internet sites, whether http or https, look into solutions that can block them from accessing sites with known malware and/or create a whitelist of sites that the devices are able to access.”

According to Mason, it is well-known that proactive security appropriate to the application is all about the packages and processes that are applied, and educating staff not to plug-in unknown USB devices, etc. But he says users in industrial environments are less accustomed to proactive security because availability is everything. “We frequently find all the networks with username ‘admin’ and password as ‘password’, and it has been that way for 20 years. And the reason it has been that way is because until recently, the network has not been connected to the internet.”

He goes on to argue that an Internet of Things network doesn’t actually always have to be connected to the internet. “It does not need to use the internet as a backhaul. There are cost implications, but depending upon the criticality of the data, you might want to keep it all completely private and have no external entry point into your IoT network. Elimination of the internet as backhaul is one way of reducing risk.”

Having said that, Mason admits that the reality is the more sensors and access points you introduce to a network, the more vulnerable you make it.

As well as security, Cisco identifies several other core challenges for network teams to be vigilant about. “Deployment of IoT on cellular networks is the equivalent of now having thousands, tens of thousands, or even hundreds of thousands of devices on your monthly phone bill,” says Bui. “You need to be able to monitor for real-time data usage and adherence to the rate plans you have negotiated.”

The teams should also have the ability to detect the abnormal behaviour Bui noted above: “You need to be able to immediately detect deviations from standard behaviour because they can be indicators of fraud, security misuse or worse. So let’s say the normal behaviour of your device is that it works only in Australia, is connected to a network 24/7, and consumes about 8MB of data per month. If it suddenly starts consuming 30MB, it could be an indication that it’s been forced into a botnet.”

Another core requirement is being able to ensure that the device can connect to the network when needed to reliably exchange data, identify connectivity issues, and troubleshoot them in real-time. That means regular network performance monitoring has to be one of the top priorities – as Mason says, having the network delay your email by 30 seconds is one thing, but if a patient’s vital signs parameters are neglected for 30 seconds it could be terminal. 

He goes on to describe an IoT deployment as just “another” wireless network implementation, albeit one that is complex and with different flavours. And in his view, every wireless network “without exception”, suffers from ‘drift’. 

“It is deployed, commissioned, installed and works perfectly according to the requirements. But from day one, things begin to change. People change environments and they add applications. For instance in our experience, people who tell us they are only going to have two cameras on the network suddenly find it is poorly performing because they added six.

“IT is a combination of people, processes and technology, and it is seldom just the technology that is the problem. It is usually the people and the process surrounding that.”