DevOps for IoT: You’re kidding, right?

by Simon Crosby, on May 24, 2018 12:28:24 PM

Those of us who have spent time building cloud stacks tend to assume that what we’ve built is so cool that it must be useful for all use cases. The cloud must be the answer to all questions, and the new workflows and processes for cloud based apps must apply. Naively, we assume that the new practices of cloud-native development and application lifecycle management will apply broadly. It’s important to be pragmatic:  David Linthicum has it right: The powerful new tools of the trade developed for cloud native apps will take forever to make it to legacy edge environments. Why?

There are two dimensions to this problem: Technology, and skill sets / work practice. Let’s start with tech.

We need to distinguish between cloud-native apps, and the stuff that’s been running enterprises forever. For a cloud-native app, data is already in, or will be delivered to the cloud, and DevOps as the new, powerful application development, delivery and lifecycle management paradigm makes total sense: I stand in awe of the fantastic achievements of the Netflix crew, who now deliver a billion hours of video per week. To the extent that IoT means “consumer / mobile apps and devices”, the DevOps paradigm seems entirely appropriate – it’s become a proven way to build, run and manage cloud applications at scale. The IoT stacks from Google and AWS are built for DevOps

Legacy enterprise applications were historically the domain of IT. Some were bought, others built in-house.  Here, private and hybrid cloud discussions related to “lift and shift” - of VMs or in some case apps - onto virtualized infrastructure - is still a key focus, and DevOps is merely a mystery. The application lifecycle is dominated by traditional IT-centric management, and OS and application patching and lifecycle management are typically manual. While there is a strong push to re-factor certain classes of enterprise apps into a containerized, micro-service based architecture, this is far from the norm.  Enterprises use and control edge equipment or assets also manage the lifecycle of bespoke and embedded systems– from SCADA infrastructure to PLCs and more. These systems are proprietary, and often control assets in real-time, so the cloud is out of the picture. Typically, these systems and their data are security and privacy sensitive. Often they are managed by operations teams. Here, DevOps is simply unheard of.

The skill set and work-practice issues should be apparent because they match the kinds of tech at play: Outside cloud-native use cases the new skill sets and workflows of DevOps are unknown and unlikely to be adopted. 

Learn More

Learn how SWIM EDX uses edge intelligence to deliver real-time operations insights from the dark data generated by IoT applications and other connected systems.

Topics:Edge ComputingIndustrial IOTSWIM AIdevops

Comments