A battle for control over the layer of abstraction that will enable services to be delivered to the network edge is now being waged between various types of hypervisors and container technologies.
Various types of embedded systems that make up a diverse number of Internet of Things (IoT) endpoints all run different types of processors. To make it efficient to deliver services to those devices there will need to be a common layer of abstraction. Given the scale of such projects open source virtualization technologies have the inside track simply because the licensing fees associated with commercial hypervisor technologies from VMware and Microsoft would prove cost prohibitive.
The choice most organizations have made historically when it comes to selecting an open source hypervisor has been between Kernel-based Virtualization Machines (KVM) and Xen hypervisors. Today Xen hypervisors are the most widely used in the cloud. But KVM has been gaining ground thanks to a recent decision by Amazon Web Services (AWS) to host new application workloads on KVM.
To make Xen hypervisors more appealing at the IoT edge, the Xen Project, which is hosted by the Linux Foundation, has been rearchitecting the Xen hypervisor to require less code that is not only less complex to maintain and secure, but also winds up delivering higher performance.
Lars Kurth, chairperson of the Xen Project Advisory Board, says the need to shrink the Xen hypervisor was originally driven by security vendors that wanted a hypervisor they could embed in an appliance. But now Xen hypervisors are finding their way into a range of embedded systems such as connected cars.
“We’re trying to take the maximum amount of code out of the hypervisor,” says Kurth.
The network edge needs more abstraction layers
The issue that managed service providers (MSPs) need to track is that abstraction layers now encompass more than just various types of hypervisors. Containers such as Docker have emerged as an abstraction layer to make it simpler to deploy application workloads anywhere, including IoT endpoints. Today most container applications are deployed on top of virtual machines. But as container orchestration technologies such as Kubernetes continue to mature, it is probable many containers will be deployed on some form of bare-metal.
To makes matters more interesting, there are other efforts underway to shrink the core hypervisor. The OpenStack Foundation, for example, is working on a Kata project that promises to converge hypervisor and container technologies.
IoT endpoints attached to the network edge represent a massive opportunity for MSPs. But without some common layer of abstraction where all those endpoints can be managed, the cost of providing managed services out on the network edge is likely to be too costly to deliver. MSPs would be well-advised to start participating more in the various open source projects that will ultimately determine how services are delivered to the edge of the network. Today most of the contributors to these projects comes from vendors. MSPs that have real-work networking experience could deliver some invaluable insights.
In the meantime, MSPs should keep a close eye on how abstraction layers are evolving at the edge of the network. They arguably have potentially millions of dollars in future services contracts riding on its outcome.
Photo: Travel mania / Shutterstock.
