One of the most fundamental requirements of any successful large or small IoT implementation is an effective application enablement platform. An application enablement platform (AEP) is a technology-centric offering optimized to deliver a best-of-breed, industry-agnostic, extensible middleware core for building a set of interconnected or independent IoT solutions for customers. The AEP forms the heart of an IoT solution, enabling customers to develop, deploy, operate and extend their IoT implementation. An AEP provides a common layer of horizontal services including device management, data management, device integration, platform APIs and SDKs, administrator and operator interfaces, monitoring and notification services, and scalability. The AEP is the foundation upon which all of the IoT components are built.
While many vendors offer effective AEPs, there are three weaknesses common to nearly all AEPs: lack of platform containerization and microservice usage; lack of IoT marketplaces; and lightness of on-platform analytics.
Weakness #1: AEPs lack a microservice-based, containerized architecture. A microservice-based containerized AEP offers many significant advantages over traditional monolithic applications. In the past, applications were typically deployed as single monolithic structures. In this context, an application is defined as the entirety of code executing on a single local machine. Given the real-world requirements of highly available and performant applications as well as the broader industry shift towards horizontal scaling, modern applications are designed with a more modular and flexible design. These modern applications run within hardware-agnostic containers and provide a collection of integrated microservices which together act as a single application. Individual microservices can exist within separate containers, separate physical machines, or even separate geographic regions.
One major advantage of the containerized microservice model is the ability to selectively scale only the the specific platform elements exhibiting performance bottlenecks. For example, while many operations staff will be familiar with the independent performance scaling required for an application service as compared to a database service, an IoT platform may present more difficult scalability challenges. An AEP solution may present a bottleneck only within the event processing module, while other application or database elements remain within expected performance envelopes. In a microservice based AEP architecture, more instances of the event processing service can be quickly provisioned on applicable compute resources while leaving the rest of the AEP solution unmodified. This enables faster, less complex and more reliable performance scaling.
Another advantage is the ability to easily deploy an AEP solution across a mix of on-premises and on-cloud environments. In particular, with the rising popularity of container technologies like Docker in the public-cloud sector, a containerized application can often be deployed by simply asking a target machine instance to grab the relevant docker image from a repository with little additional configuration required. Containerization also prevents public cloud vendor lock-in, as all major public clouds support technologies such as Docker. Thus it becomes a far simpler task to port an IoT deployment from one public cloud vendor to another as corporate relationships, price incentives or technological advancements dictate. Customers employing AEPs with legacy monolithic designs may need to invest in complex application profiling and debugging to locate underperforming platform elements, and once identified, such monolithic applications may require a fresh instance of the entire AEP just to support a single service requiring more compute resources. MachNation believes that all AEPs must implement these technologies in order to offer effective solution deployment and flexible scaling models.
Weakness #2: AEPs lack an effective IoT marketplace. An effective marketplace provides customers with fully productized and solution-specific offerings, saving significantly development effort and shortening time-to-market both for initial solution implementation and iterative functionality improvements. IoT marketplaces. An IoT marketplace provides a customer-facing online store. These stores should offer platform compatible IoT hardware such as gateways, modules and development kits. They should also offer software solutions such as pre integrated services, devices, protocol compatibility agents and even entire vertical solution software stacks.
While IoT application and hardware marketplaces are relatively new to the AEP space, their ability to enable customer success is paramount. By choosing from a catalog of existing hardware, services, solutions or applications, IoT marketplaces offer customers the flexibility to pick the best implementation path for a given set of solution requirements. While some AEP vendors prefer to leverage in-house development teams to provide customers with these choices, this ultimately provides a more opaque process for customers, typically resulting in more costly solution implementation.
Weakness #3: AEPs lack strong on-platform analytics capabilities. MachNation has identified two specific categories of IoT analytic offerings: descriptive analytics and predictive analytics.Descriptive analytics emphasizes visualization and monitoring of both the current and historical state of a given set of observations. In contrast, predictive analytics emphasizes the construction and maintenance of data models enabling various process optimization capabilities. While nearly all AEPs offer some form of analytic insight, often these capabilities are limited to descriptive analytics such as graphs of real-time or historical data and rarely include features such as regression analysis or other basic model building tools.
While this more limited feature-set is often sufficient for a high-level overview of “what is going on” it does not consider “what will happen” or “how to make a given process happen more efficiently”. Given the potentially deep wells of data provided by IoT solutions, customers, especially small or medium sized enterprises, may not have the dedicated data science resources capable of leveraging these novel data sources. As a result, many customers leave potentially valuable optimizations or insights undiscovered and unexploited as they lack the ability to derive business-value from the deluge of ingested IoT data. Therefore, MachNation believes that a highly effective AEP should provide both descriptive and predictive analytics methods on-platform to enable customer to best generate business value from their IoT deployments. Monitoring views and basic representations of historical data alone do not constitute a fully-effective solution.
An AEP is a fundamental component of a successful IoT solution. While some customers may attempt to build their own implementations, this is often a costly and complicating mistake. Instead, choosing a fully featured AEP offering that provides these key functionalities out-of-the-box leads to a less costly and more effective IoT implementation.