Strategy for scaling IoT: How to plan for the real world

Scaling of IoT is an extremely important aspect that shouldn’t be overlooked as typical scale scenarios see devices run into thousands and millions of devices, exponential growth rates are to be accepted as a reality in design, setup, and operations.

By Yogendran Gopalakrishnan

As the Internet of Things (IoT) is becoming a mainstream part of human life, there have been many challenges that have started cropping up. It has posed very fundamental challenges to industry concepts such as scaling, robustness, consistency and not to mention – the old tried and tested industry techniques of planning, forecasting, upgrade cycles and economy of scale. Disruptive technologies with their unconventional characteristics demand unconventional approaches towards planning, scaling and engineering.

Scaling of IoT is an extremely important aspect that shouldn’t be overlooked as typical scale scenarios see devices run into thousands and millions of devices, exponential growth rates are to be accepted as a reality in design, setup, and operations. An immediate statistic which brings to bear the humongous challenge posed by IoT in the area of scaling is the very fact that there are to be an estimated 20 billion connected IoT devices by 2020 Here are some recommendations on how organizations should strategize and plan on deploying IoT at a sustainable scale in order to derive the maximum business benefit:

Have the Cloud and Big Data built into your strategy from day one – Countless books, papers, articles and PHD theses have gone into how a cloud first strategy brings scale and cost predictability. In essence, cloud brings with it quicker time to market, rapid scale up and down, linear and predictable pricing models and last but not the least – the robustness and architectural patterns for scale that IoT is verily an immediate candidate for. It is only a matter of time before volumes and scales kick in which make big data compliant architectures a pressing need to plan for from the scratch.

Don’t pass on “PaaS” – Several tried and tested PaaS solutions have emerged in the IoT space especially by the big cloud players. It is exciting times indeed for the IoT space as vendors jostle for space in what is now a crowded market, each vying to catch customer attention with PaaS features and capabilities spanning the entire spectrum across device on-boarding, device configuration, device data ingestion and analytics; all delivered following the best patterns and scale built in from scratch.

Cloud and EDGE, the potent combination IoT needs – It is by now wide spread knowledge amongst IT executive circles that “on premise” and other custom infrastructure approaches will never reach scales and the cost optimizations that cloud readily promises from day one. That doesn’t mean that all processing work needs to be shipped off to the cloud and that is where EDGE resources and infrastructure becomes a critical player in the game. EDGE devices are part of a cloud computing paradigm wherein processing not happening on the cloud typically happen on the final layers of the network where the data is actually generated. In many a scenario involving immediate and speedy processing of data, some local processing of data in the form of collation and filtering at the EDGE before being dispatched to the cloud works out faster, cheaper and brings in predictability to boot.
 
Keep an eye on the first and last Mile in your entire flow chain – The chain is only as strong as its weakest link. EDGE and cloud infrastructure means that the first and last mile are already on to proven patterns, but it is not always possible to make sure they figure in your setup. Some IoT scenarios also involve co-existence of EDGE alongside devices which talk directly to the cloud or external processing systems. But a slow security processing system for your devices, a hardware interface which cannot keep pace with the speed needed, poor mobile signal network strength etc. can delay the process. Many a cloud processing scenario also involve two-way communication between the cloud and your connected device so “keeping up” with the strongest parts of your link becomes important.

Plan and leg room for your traffic peaks – Many IoT scenarios involve several devices “synching” with the cloud in specific windows of time or whenever they come online which is a typical scheduled operation. Peak scenarios in IoT are far more complex to plan for and predict than their typical counterparts from other scenarios, say in banking or retail. Some PaaS platforms offer a capability called “auto scaling” wherein more critical resources are made available during peak scenarios against “pay as you use” billing plans. Hence best practices and patterns like auto scaling up and down of resources like containers, CPU cores, RAM, disk space etc. are must haves in a PaaS setup for IoT scenarios. Integrate peak scenarios in your application development cycles and validate peak scenarios specifically to ensure the “auto scaling” capability is able to keep pace with the traffic.

Get your Application Architecture right – Bad application architecture and design is antithetical to IoT scaling and you stand to negate all the advantages accrued from the patterns and best practices of the underlying cloud PaaS. The application architecture must be able to scale up and down to incoming traffic keeping pace with the resources made available to by the underlying PaaS platform. Scaling up and down shouldn’t necessitate code changes or application deployments but more achieved by tweaks to configuration. Plan to keep your architecture open to the latest and greatest the emerging industry trends have to offer in terms of extension but not modification.

It is very common today to hear about startups see their IoT products succeed and sell by the millions. On the other end of the spectrum, we see large and well-established industries and enterprises take their IT infrastructure to the next level by having every phase of their manufacturing and supply chain retrofitted with millions of IoT devices. Hence, scale planning is equally applicable to every player across the size spectrum and not only restricted to a few players. Business cases involving connected devices sharply increasing from the low thousands to several millions are not uncommon. In the face of such enormous opportunities and the uncertainty using the above potent mix of planning and engineering practices will ensure scale is achievable in a cost efficient and linear manner.

The author is Big Data & IoT Architect, SAP Labs India