The latest insights from your peers on the latest in Enterprise IT, straight to your inbox.
This article is by Featured Blogger Nitin Kumar from his Blog Page. Republished with the author’s permission.
IoT is now widely regarded as a transformative force. According to Juniper Research, by the end of 2020, there will be 38.5 billion devices connected to the internet — a growth rate of 285% since 2015.
But as more devices interconnect, the number of malfunctions will only increase. This is an unresolved issue, even outside of security. For example, based on the data from Juniper, even if 2-3% of devices were to have performance issues, that would be 500 to 750 million devices.
The Juniper report notes that users have reported at least 1.2 problems daily. These scales will overwhelm IT support and operations teams. Finding a problem quickly and trying to fix it faster just won’t work, and simple automation tends to break down.
According to a Vanson Bourne survey (via IoT Now), roughly 3 out of 4 CIOs fear IoT performance problems "could derail ops and hurt revenues."
After working in the IoT space for the past five years and interacting with several players across the spectrum, I have observed a few reasons for operational failures within IoT. Here are six operational challenges that come with IoT:
1. Granular Configurations
The edge has many granular configurations that are personalized and connected via APIs. These often require manual settings, configurations and interventions.
2. Network Limitations
In the era of sensor- and machine-generated data, the size, complexity and attributes of the workload will attain massive scales. In the absence of large-scale edge network implementations today, incoming workloads can overwhelm devices, and network limitations create performance problems at scale. A high density of IoT devices also increases network congestion. There is also lack of presence detection in certain realms of IoT where everything must traverse through a smart hub or router. As a result, logging, monitoring, reporting and other operational functions will quickly get beyond humans.
3. Workload Issues
Network limitations and bandwidth constraints are constantly on the rise. The proliferation of more devices adds to the load. Many IoT features require lower latency to effectively use and will require either local servers or service providers to provision new bandwidth and QoS for workloads with unique requirements. This will become costly to implement even without factoring in the human cost of operations.
4. Environmental Factors
The physical environments where IoT devices are located might include high humidity, extreme temperature, untested terrains, etc. These variables impact IoT device performance in myriad unpredictable ways. Operations needs to be prepared to prevent, detect and fix issues autonomously without human intervention at certain scales.
5. Integration Problems
Many devices require their own software, which may not necessarily integrate with standard IoT gateways, hubs, routers, protocols, etc. The lack of integration between various interfaces will create higher rates of failure and longer detection times.
6. Fragmented Service And Support
The ability to quickly detect, analyze and repair problems can determine the success and scale of IoT devices. OEM warranties are expensive, and there is a shortage of IoT skills to operate billions of devices. ISPs and SIs have more responsibility for the performance and operations of wide-ranging complex devices while dealing with a learning curve and skill shortages. Company IT departments are overwhelmed with thousands of new devices to support and millions of alerts coming in on a regular basis, making detection, prediction and fixing that much harder.
How To Help Scale IoT
The implementation of edge computing could address some concerns with data management, latency and network reliability. But keep in mind it will still not be able to offset the operational complexities outlined above. The problems attributed to remote survivability, large datasets, rapidly changing workloads and hardware integrations will still exist.
Autonomous capability senses the state of a system in relation to the context and makes decisions and performs actions based on both context and state. The context is not only the environment, but also the ideal state of external factors such as workloads, events occurring in the ecosystem, etc.
Autonomous operations would need to be operated with very different metrics in mind. Gone are the days of traditional mean time to resolution (MTTR) at the scales in question here. Autonomous operations can measure mean time to prevention (MTTP) and mean time to identification (MTTI).
It is recommended that organizations get ready for autonomization and start to develop their operational strategies by thinking about their use cases and workloads, and the availability of skills to resolve issues at a wide variety of locations. While automation might be a great start, autonomization is the ultimate destination for IT operations.
In conclusion, IoT-driven workloads at very large scales make it challenging to manage in the traditional manner. The complexity and variability cannot be overcome manually or even with basic levels of automation. If any of these operational challenges creep up, make sure to consider the advice mentioned above to help scale your IoT-driven workloads.