How to Overcome the Challenges Facing M2M Operators
Posted in Blog, M2M, Wireless Operators, Network Management, Service & Network Management,Data Analytics, Applications, Performance Monitoring, Service Providers, Billing, Provisioning, Security,Wireless
By John Molise
The exploding marketplace for machine-to-machine (M2M) communications and the Internet of Things (IoT) is creating a great deal of excitement. The difficulties around successfully operating networks and systems in IoT/M2M services are often overlooked, since every deployment seems to bring a new set of problems. Here are some of our top challenges:
Lack of visibility and information: Missing from most IoT programs are key elements of traditional network support and troubleshooting, namely a live person on the other end of the line who can provide valuable input about user experience and the local conditions of the remote device. IoT programs involve tens of thousands to millions of devices that are managed centrally by a single person or entity. Often, information as fundamental as the location or street address of the remote device are unavailable – let alone empirical data such as RF conditions, device errors or application performance criteria.
The typical support case involves customer input such as, “My application is supposed to be doing X, Y and Z … and it is not doing it." That leaves full accountability for triage and analysis on the technical support team. With traditional network appliances and instrumentation, the team would be blind. Traditional network logs do not show the complete picture, as network connectivity problems are often represented in system logs by an absence of activity (as opposed to some positive confirmation of a problem like an error or fault code).
Dynamic instrumentation and monitoring are critical to the successful management of IoT programs. In our business, we keep everything, no matter how inconsequential the data point may seem. For the support case described here, M2M/IoT providers need to understand whether the device was doing “X, Y and Z" at the same time yesterday, one week ago today, or one month ago today. They need to see how that activity was depicted through every last drop of available cellular signaling, control plane and user plane traffic data and determine the variance between that profile and current conditions. Data and data analytics mean everything to successful IoT operations.
Variable use of network transport: A subscriber base of 50 million retail consumers will generally leverage a mobile network for one common use case. That is a network management and support luxury not afforded to operators with IoT programs. Ten million connected devices comprised of 1,000 different enterprise programs means the operator will support 1,000 distinct use cases and workflows across this device population.
The only way to operate successfully under these conditions is to provide platform flexibility that enables a customized user experience. I am fortunate enough to be surrounded by the best engineers and developers in the business, and without access to this Silicon Valley talent, I would not be able to support the unique network capabilities, billing and rating flexibility or provisioning and supply-chain requirements that my customer base demands.
The attack of the machines: The early days of M2M – where each new customer deployment constituted a distributed denial-of-service (DDoS) attack on network and system resources – might have passed. But the fundamental characteristics of machine behavior remain.
Machines lack the graceful randomization of consumer traffic. If 50,000 humans were told to initiate the same action on the network at the same time, there would naturally be some deviation from the expected behavior. Now when 50,000 machines are programmed to do something at a time-synchronized interval, they will do that very thing, and they will be relentless in getting what they want from the network. Machines also lack the intelligence to back off when faced with unexpected conditions (whereas humans are pre-programmed with a tendency to give up out of frustration and lodge a complaint with their service provider).
All of this means the service provider must operate with a couple of principles in mind. First, the network service must be available every second of the day. In an environment where a 30-second interruption to any leg of signaling, control plane or user plane service can drive a 10x-20x increase in machine transactions, keeping services available is by far the easiest and least expensive option for the operator. Secondly, when the invariable traffic spike comes (keep in mind that an outage within the customer premise will drive the same behavior), we require the capability to tune our network and systems infrastructure to handle the event gracefully. Traditional load balancing and horizontal scalability can take us only so far – a practical limitation that is driving my own organization to push the envelope of virtualized network functions, elasticity and autoscaling.
The recent announcement of our Infinity Support program at Aeris was formal recognition that there is no easy path to success in IoT operations. Success requires constant collaboration, a best-in-class customer base who win in their respective verticals, and a team of M2M experts who pull it all together, recognizing that if we don’t get a little bit better today we will fall behind.
This has by no means been a full account of the challenges we face in operating networks and platforms for the IoT marketplace, but it highlights a common theme. To be successful, we must be willing to evolve and adapt to an ever-changing set of requirements, and our systems must be flexible and innovative to allow us to do so profitably.
John Molise is vice president of operations for Aeris Communications, responsible for day-to-day operations of the Aeris network and services platform.