Victoria - Fotolia
The Internet of Things is coming on full force, and IT has some work to do.
The fundamental idea of Internet of Things (IoT) is that connectivity is rapidly growing -- via the Internet -- to a wide range of embedded sensors, devices and systems. IoT embraces existing machine-to-machine communications and expands to include more analytics and consumer-oriented products.
By the end of 2020, there will be approximately 212 billion Internet-connected things, from healthcare instruments to tracking devices to wearables, according to analyst firm IDC.
And all of that new or expanded instrumentation generates traffic for networks and demand for storing and processing data, which is bound to mean more stress on server, storage and networking infrastructures, said Nik Rouda, senior analyst at Enterprise Strategy Group (ESG).
More devices, more disorder
Back-end systems -- servers, storage and networks -- must grow into an Internet of Things architecture to support the massive economic transition.
From an IT support perspective, the large number of small packets of data coming from multiple directions could cause chaos, said Ed Featherston, director and senior enterprise architect at Collaborative Consulting, an IT services provider based in Boston. These small packets of data come from numerous devices, ultimately consuming corporate bandwidth.
Without proper planning, "IoT could overwhelm a corporate WAN or [create] bottlenecks at remote or hosted sites," said Featherston. "Any IoT devices that have near-real-time needs should absolutely be designed and planned for, but ... since most is data capture for use by analytics, network latency may be less of an issue."
IoT will put new demands on storage, Featherston added. "These tremendous amounts of data need to go somewhere to be useful," he said.
Act now for IoT technology
First, address how you plan, build and manage your infrastructure.
"The IoT and the Internet of Everything will change the management and operation of infrastructure," Clark said. "If you operate in the traditional stovepipe manner, you will struggle to keep up, unless you have a comprehensive approach to lifecycle management and a unified management layer."
Second, understand and stay on top of traffic patterns and key data so that you can conduct analytics in an effective and less distributed manner.
And finally, plan to move back toward a more distributed approach to computing. "If you don't simplify and distribute, you will crater when IoT hits."
Pushing IT to the edge
Experts say growth alone won't be enough; a new Internet of Things style of architecture may be necessary.
The result of IoT's proliferation could be a sea change in IT, according to a recent Gartner report, "The Impact of the Internet of Things on Data Centers." Because of the massive volume of data that IoT produces, the recent trend toward centralizing applications to reduce costs and increase security may become unsustainable.
Instead, organizations must aggregate data in "multiple distributed mini-data centers," as explained in the Gartner report, where at least initial processing occurs and key data is transferred to central facilities for further processing.
"The magnitude of network connections and data associated with the IoT will accelerate a distributed data center management approach," implying a coalesced security/data/management environment, Gartner reported. This challenges organizations to manage the environment while wrestling with remote storage governance and absolute limits on what data can be stored.
"IoT creates massive scale requirements around IP address management and related services, such as domain name system and dynamic host configuration protocol," the Gartner analysts said.
IT must consider connecting all of the access layers, or edges, of the enterprise, said Jason Beiter, enterprise solutions architect at Annese & Associates Inc., a managed services provider specializing in networking and IT infrastructure. "This essentially means pushing out that connectivity from the core to the edge," he said.
But the nature of IoT traffic provides a reprieve to data center architects. Most machine-to-machine traffic is local, as clusters of devices communicate with one another using something other than Wi-Fi or cellular, such as Bluetooth, radio frequency identification and near field communication, said Azmi Jafarey, CIO at Ipswitch, a file-transfer software provider. Those clusters aggregate data onto one or more devices that connect to the Internet.
For IT, the issues are potential interference, more data traffic, and security. As bring your own device policies are necessary to protect corporate data, IoT policies are necessary as wearables -- more smart machines connecting to the network -- come into your offices, Jafarey said.
Strategically, the scale and nature of the load problems caused by IoT technologies on traditional architectures vary widely, depending on the industry and the specific nature of an application, necessitating different approaches, said Paul Brody, IBM vice president and North American leader for mobile and IoT. For example, a single jet engine can generate multiple terabytes of data on a flight, challenging an IT organization to process, store and make sense of that data for an airline to optimize business. Because the data holds the key to the efficient operation of a machine costing tens of millions of dollars, it will probably be worthwhile, Brody said.
On the other hand, he said, "Our studies show a scaling problem in terms of mid- and lower-range IoT applications, where the very high levels of data volume do not match the potential value of the information."
Much of the information from IoT will likely be compressible because most of the data points may be unchanged, said ESG's Rouda.
"I expect to see a lot more aggregation into metadata and local/edge-based analysis taking place, rather than shipping every data point back to the data center," he said. "Data will also need to be encrypted at the edge."
But as data aggregates, machine-to-machine traffic will hungrily consume wide area network (WAN) bandwidth. And local access network (LAN) traffic can also increase with a poor design in data extracting, transforming and loading between systems.
"Latency could also become as much of an issue as bandwidth, particularly for real-time use cases requiring rapid responses," Rouda said.
Planning and engineering an IoT architecture can help. For example, IoT devices should operate over nondedicated, preexisting networks, so sensors should be engineered to treat their supporting networks as "best effort," said Kent Sanders, enterprise architect at Tata Consultancy Services, a global IT consulting company. The sensors should be able to buffer nondeliverable data until a later time, and have flexible retry or be able to discover alternate data delivery channels, he said.
"Typical network segments are not designed to connect very large numbers of devices -- this will cause contention issues with shared-media networks, like wireless or fan-out issues with switched networks like typical Ethernet," Sanders said. "So, you should ask yourself, 'What would happen if I needed to add 300 permanent, low-bandwidth stations to my wireless LAN?'"
Companies should ramp up their ability to handle large volumes of data, said Eric Herzog, chief marketing officer and senior vice president of alliances at Violin Memory Inc. "There are software companies like Splunk that deal with real-time processing of IoT-type data sets, so you want to get the right software in place," he said. Then, be sure to look at your server infrastructure and at potential gaps in I/O between servers and storage infrastructure.
The future of IoT probably involves an architecture that pushes the application layer to the router and integrates a container with logic. This way, the network doesn't have to shoulder the whole burden, said Mike Martin, CIO of nfrastructure, a systems integrator in Clifton Park, N.J. "Otherwise, there is simply no way to add enough bandwidth."
The ubiquitous and growing power of the cloud may provide a partial answer.
Cloud architectures will be the key enabler of industrial IoT technologies due to the cost benefits, scalability and elasticity, said Walid Negm, global R&D director at Accenture Technology Labs and lead of Accenture's Industrial Internet Strategic Innovation Initiative. "We see the cloud as a more prevalent paradigm than the edge," he said.
However, there are some scenarios where the edge compute model is preferred. For example, low latency matters in certain applications. When network delays will not hamper the solution, centralized hosting wins.
"There is emerging thinking that criteria for edginess need to be created, including bandwidth and latency. For any given use case ... there are a range of functions that must be performed," said Negm. "In principle, you have to choose whether the function is performed centrally or at the edge."
How an Azure model can help with IoT development
IT priorities 2017: Bigger 2017 budgets enable more ambitious IT activities
Dig Deeper on Internet of Things (IoT) Network Infrastructure
IoT security regulations refocus data privacy best practices
Technology risks: What CIOs should know and steps they can take
Ship data to provider, switch on machine learning capabilities
Tips for CIOs: How to implement machine learning projects