As the “Internet of Things” grows, mobile phones are becoming personal controllers that connect consumers through Bluetooth, Wi-Fi, ZigBee and other personal area networks to home control – cameras, smart watches, glasses, fitness trackers and even smart clothing. Because of that connection, the volume of mobile interactions involving mobile apps and new IoT devices is growing, creating a challenge for those who want to monitor how IoT devices and their applications perform in the real world.
Many of the projected 25 billion interconnected “things” in 2020 will be accessed and controlled by mobile smartphones through multiple applications and over PANs. Both operators and manufacturers increasingly want to understand how actual consumers experience the performance of these devices, whether they are smartphones, wearables tethered to smartphones, or other devices controlled by apps and connected over unlicensed spectrum. Unfortunately, even when using big data technology, the infrastructure required to securely track and monitor this mountain of telemetry becomes overburdened if not designed for scale. Business stakeholders must perform data collection in a way that minimizes the amount of downstream processing.
The data types to be sourced from devices vary widely and depend upon each specific set of stakeholder requirements. For example, network teams may want to track voice-over-LTE or Wi-Fi call drops. Supply chain teams might be interested in device power metrics associated with a tethered smartwatch model or the degree to which Bluetooth or other connections drain power. A customer care team might want to understand whether a connectivity issue is a device or network problem, or if some specific connected devices create performance issues; while a marketing group could be more concerned with how overall consumer device performance impacts churn. Each use case requires specific metrics that are often sourced by different stakeholders through independent means, relying on aggregation into data lakes for downstream processing. As a result, when the number of deployed devices grows, cleaning, filtering and reporting across this wide and varied data set quickly becomes costly, both in terms of time and compute resources.
Consider the vast size of this challenge: Gartner estimates that the number of device-resident services that track mobile activities will reach 100 per day by 2017. If a mobile operator grows to deploy 20 million or 30 million smartphones and wants to understand consumer experience over time and geolocation, they might collect 50 kilobytes of smartphone telemetry per day and another 5 to 10 KB of telemetry for each tethered device or mobile service tracked. This means that the total telemetry collected in just one month could easily reach 1 petabyte or more.
Monitoring actual mobile customer experience requires that a software agent be deployed onto mobile devices. To maximize scalability while serving the needs of multiple stakeholders, including the privacy needs of consumers, it is imperative that these agents be used not just as bulk collectors, but as centrally controlled processors with integrated privacy and security mechanisms that allow for secure collection, preprocessing, filtering, aggregating and encrypting telemetry prior to transmission. Some IoT devices have their own CPU, operating system and storage so that a small level of distributed processing and encrypted storage can occur directly on each device. Other classes of devices, particularly those that may be tethered as a slave to a mobile phone master, may afford little or no ability to locally process and store aggregated data. In that case, the mobile phone can be used to perform preprocessing functions. In all cases, centrally controlling the computational loading across field-deployed agents allows processing to be distributed to optimize how information is collected, maximize security and minimize the impact of transmission on the network.
Collection, of course, is just the first challenge at IoT scale. To quickly gain insights from collected device metrics, domain-specific analytics tools should operate in concert and integrate with device collection agents so that the flow of metrics received from devices is streamlined and secured for each analytics transaction. This approach seems contrary to the perceived value of big data lakes, where relatively unstructured data is combined from multiple sources and made available for later use.
In a July 2014 article, Gartner points out that simply dumping collected information from multiple data sources into a single data lake creates a perception of lower cost and efficiency because IT no longer needs to understand how the data is to be used.
“The burden of getting value from the data then falls onto the business user,” Gartner stated.
When telemetry from disparate mobile device data sources is simply aggregated into data lakes, this apparently efficient collection process results in downstream data processing that becomes unmanageable at scale. The result is that business users are not able to gain insights from data that they know they already have. The time and cost of processing and reporting downstream simply becomes too large.
To overcome these issues, operators should consider implementing solutions for mobile intelligence gathering that integrate a massively distributed device collection mechanism with purpose-built analytics. This will help assure that downstream performance remains manageable as the number of devices, applications and the additional devices controlled by them grow. Such a solution should deliver three key capabilities:
1. It should abstract the underlying collection mechanisms so they can be centrally controlled in real time to achieve different goals without having to redeploy individual device agents in the field.
2. It should automatically integrate the device collection process with purpose-built analytics modules so that collection is optimized for the analytics to be performed.
3. It should allow vertical domain experts to add analytics modules over time to support the ongoing changes in data, voice, application, power and stability metrics to be analyzed, but without requiring them to be experts in the underlying big data, database, messaging, privacy and device-specific software technologies involved.
For over 25 years in the information technology and communications industries, Michael Tanner has been a leader in bringing game-changing technology products and services to market. Prior to joining Carrier IQ, Tanner was CEO at Bamboo Solutions, a collaboration software company, and CSO at Edge Communications, a provider of voice, data and IT infrastructure solutions. He has served as a board member for five technology businesses ranging from startups to more than $100 million in sales, and has served as a guest lecturer at the Haas School of Business at UC Berkeley, Santa Clara University and the Kauffman Foundation on topics of market adoption and technology market strategy. Tanner’s early career included roles in sales, product management, marketing and business development at Schlumberger, Autodesk and two successful engineering automation software companies.
Editor’s Note: In an attempt to broaden our interaction with our readers we have created this Reader Forum for those with something meaningful to say to the wireless industry. We want to keep this as open as possible, but we maintain some editorial control to keep it free of commercials or attacks. Please send along submissions for this section to our editors at: [email protected].
