For a consumer, a network issue can be annoying and inconvenient, perhaps requiring a re-boot of the home router or a call to customer support, but for mission critical Internet of Things (IoT) networks a network issue can be considerably more serious. For example, networks running manufacturing processes or oil and gas drilling can be extremely costly if they fail, even for a short period of time and failure of networks in hospitals can be potentially life threatening. It makes sense therefore for these types of networks to look for ways to make them more secure, more reliable and better equipped to deliver the performance needed by each individual IoT end-point at a reasonable cost.
According to Gartner, over 60 percent of network issues are Wi-Fi related and it is clear that a vast number of IoT devices are now and will be connected over Wi-Fi, so solving at least this part of the network would dramatically improve the overall performance of the network. One way to accomplish this would be to hire a team of Tier 2 Wi-Fi technical specialists who could constantly monitor every access point and perhaps even try to look at some of the devices connected to each access point from time to time. This is clearly not a practical solution and certainly does not meet one of the key requirements which is that this monitoring must be done at a reasonable cost. A better, more automated solution is to use a cloud-based, big data environment running an artificial intelligence (AI) software suite that automatically monitors, manages and optimizes the Wi-Fi, focusing on quality of experience for IoT devices and end users by delivering proactive monitoring, detection, self-optimization, mitigations and recommendations.
Leveraging the benefits of AI to simultaneously monitor in real time, all interactions with all devices attached to every Wi-Fi access point can be done if the solution runs in the cloud within a big data environment as the data gathered can be substantial over time. The low-level radio and network data gathered can be analyzed and utilized to optimize the end user experience or IoT connection, dynamically adapting the access point in real time where possible or by providing an alert with a clear, actionable root cause, thereby significantly increasing the efficiency of IT resources. AI coupled with machine learning (ML) also enables the network to ‘learn’ the environment and proactively improve service levels. Some of the key quality of experience (QoE) parameters monitored by this solution could include:
- Intuitive visualization of current and historical device QoE state for each location.
- Historical recording of all device KPIs and WLAN attachment events for each location.
- Auto-correlation of KPIs grouped in common device QoE root causes.
- Analysis of actual 802.11 physical layer data rates behavior for each device. Deviation from expected values given measured site RF parameters.
- Analysis of Device and AP rate adaption abnormalities causing spread of data rate selection.
- Analysis of Wi-Fi frame integrity failure and impact on actual available traffic capacity to each client devices at any given time.
- Analysis of dynamic sharing of channels by the onsite devices, neighbor WLAN networks and non-Wi-Fi devices.
- Analysis of coverage issues affecting end users.
- Analysis of temporal noise impairments affecting end users.
- Analysis of hidden node situation causing client link impairments.
- Analysis of authentication failures and specific Wi-Fi capabilities mismatch.
- Misalignment of device firmware baseline.
Another key attribute to consider is that not all IoT devices are equal, leading to the concept of “device profiles” which enable the IT staff to classify IoT devices based on the performance they need to function as designed. For example, a video camera requires high bandwidth but if it occasionally disconnects for a few seconds that may be acceptable; conversely a sensor in a delicate manufacturing process may not need much bandwidth but it has to be connected 100 percent of the time. Monitoring and managing the access points to provide the performance needed by these different types of devices gives us the concept of optimizing the QoE for each device type based upon its profile.
Finally, another benefit of a big data, cloud-based solution is that connection and performance data can be stored for a long period of time which means that the system holds detailed historical data thereby eliminating the need to recreate problems in order to find the root cause. As Gartner has noted and almost any IT person will affirm, finding the root cause is about 90 percent of the time to fix the problem so not having to recreate a problem dramatically saves IT resources.
In summary, the benefits of an AI driven network can be very significant if the Wi-Fi IoT network is large enough in scale with mission critical requirements. All users and devices would see improvements in QoE, in some cases dramatic improvements, without utilizing valuable human IT resources.
About the Author:
Huw Rees is a Business Development Advisor at KodaCloud. Prior to joining the company, Huw held several senior leadership roles at 8×8, Inc., includingvice president of business development, VP of Sales & Marketing and CEO of Centile, Inc., a subsidiary of 8×8. As an officer and senior member of the 8×8 management team for 16 years, Huw was responsible for the execution of the company’s shift to a Unified Communications as a Service (UCaaS). Huw holds several patents in the cloud PBX technology field. Rees received a B.Sc. (Hons) from the University of Manchester (England), Institute of Science and Technology in Electrical and Electronic Engineering and has an MBA from the University of LaVerne (USA).