• One of the greatest challenges with IoT isn’t device instrumentation or even data storage or analysis. It’s integration and how you move instrumented data (at speed) between endpoint, edge device, gateway, processing engine, data lake and analytics software.
• This focus on data integration coupled with the emergence of cloud-born software development/deployment practices will lead to a resurgence among traditional middleware vendors TIBCO, Software AG and Red Hat.
Memory deceives us so gently sometimes, like an old friend whispering in our ear, telling us that what has gone was so much nicer than what we have now. For me, I miss my childhood friends and home, my days at college, and most certainly my clear case Apple Newton. My recollections of those times and artifacts are so real, so warm and reassuring. And of course they’re each an absolute lie, as proven time and again by scientific research. We create the past anew each time we draw a memory to the forefront of our attention.
I look back on the heyday of middleware and of service oriented architecture (SOA) in particular, therefore, with a mixed sense of both nostalgia and distrust manifested in a rosey longing for software that could pull together and expertly orchestrate any and all systems of record and engagement. It should have worked, just as we should all be flying around in jet cars by now, I suppose. But the grand vision of SOA which dominated so many top down IT discussions back in the ’90s has died away and was replaced by something a more sensible and achievable in the form of basic data and application integration.
But the ideals of SOA live on in today’s integration solutions and guess what? They’re about to see a dramatic resurgence in stature and value, driven principally by our industry-wide fascination with the Internet of things (IoT). So much of the energy expended by IoT technology providers to this point has focused on the instrumentation of devices and the security, storage and processing of data, that the basic challenge of integrating disparate data sources at speed has only been addressed as a component of big data technologies like Hadoop and Spark. We’ve ended up with a two-tier architecture between device and cloud, which unfortunately cannot accommodate or exploit the huge volume of data and data sources involved in any sizable IoT deployment.
What we need therefore are vendors focused on what happens between the instrumented device and back end data processing/storage. In short, we need more middleware. Fortunately, what were once known as SOA vendors Red Hat, Software AG, TIBCO and others are actively exploring that technology gap. These players are not just building data processing at the edge (as we’ve seen from leaders like Cisco with its data virtualization and FOG processing) but also tackling the tough challenges of data marshalling and transaction messaging between various IoT constituencies (device, gateway, LoB system, etc.).
These players are building on their historic middleware foundations capitalizing upon modern, synergistic ideas like microservices and flow-based programming, the manifest value of open source software (OSS), and scalable platforms like Amazon Web Services and Microsoft Azure. The results so far are very interesting. There’s Software AG’s recently introduced IoT Analytics Kit for Apama Community Edition, which basically pushes the company’s streaming analytics capabilities down onto devices like the Raspberry Pi. And there’s TIBCO’s recently introduced Project FLOGO. This OSS development framework literally pushes middleware down onto the IoT device itself, allowing developers to build IoT apps that is tolerant of network disconnects, conscientious of communications costs and available storage, able to engage in on-device or at-edge decision making, all while applying proven integration capabilities to the problem of securely connecting those devices with gateways and back end systems.
Do I think efforts such as these will usher in a second coming for SOA or even a market resurgence around middleware itself? No. I’m nostalgic, not wistful. But I do think that these middleware players will play an influential role as IoT unfolds, showing both IT buyers and big data players that sometimes the old ways are indeed the best ways.