VMware Moves Into Edge Computing

One of the most exciting revelations last week at VMworld 2018 was VMware’s futuristic vision of delivering cloud to IoT devices, or more generally speaking, to the so-called Edge devices.

Edge devices, connected to the internet, are widely predicted to be ubiquitous in the near future. The market for edge devices is estimated to grow 30% compounded annually for the next 3-5 years, and to be worth upwards of $6 billion by 2022. This is a serious opportunity, and there are several big players out there trying to snatch a piece of the pie.

Challenges of Edge applications

There are a few reasons why modern machine learning applications in the Edge, like robotics, smart homes, and self-driving cars, can’t offload all of their compute-intensive processing to the cloud.

Massive scale of generated data

Most edge devices consist of, at the very least, sensors that measure some property of the real-world and generate unstructured data. This is a huge amount of data per second – think of a set of security cameras in a football field that are attempting to model and identify suspicious activity. Uploading all of this data to the cloud for processing would be very expensive in terms of bandwidth, among other costs.

Real-time decision making

Some edge devices, like sensors for agricultural farmlands, are only capable of measuring and monitoring. But there are a huge number of applications in which the device needs to make real-time decisions and take action. These applications would be negatively impacted by the network latency cost of uploading data to the cloud for decision making.

Intermittent connectivity to cloud

The device needs to be capable of functioning independently, without access to the internet and the cloud. A smart car can be expected to drive through areas without a decent internet connection, but can’t afford to stop functioning while on the road.

Project Dimension

VMware is working on a project that is aimed at providing a seamless, unified approach to solving the Edge problem.

Integrating the Edge into the multi-cloud story

With VMware Cloud on AWS and vSphere on their customers’ on-premises data centers, they have already made headway into providing a single-pane-of-glass interface and management plane into a customer’s multi-cloud environment. Customers can now simply log into their VMC consoles and get access to all of their software-defined data centers (SDDCs), they can monitor health, manage resource allocation, deploy hosts, and do a variety of things with the operational consistency of vSphere that they know and love. Project Dimension takes this a step further, integrating edge devices into this “cloud mesh”. The aim is to extend the same operational consistency and compatibility that already exists between customers’ on-premises SDDCs and AWS, to their edge devices, which would be exposed as virtualized nodes in a cloud of their own.

Hardware-as-a-Service

VMware plans to roll this out as not just software, but also with accompanying hardware at the edge. The hardware would be hand delivered and set up by VMware’s technicians. It will automatically connect to VMware Cloud, which will manage deployment of all infrastructure software updates and firmware upgrades. This hyper-converged infrastructure will not only make the edge device highly optimized, but also make it easier for trouble-shooting, enabling VMware engineers and support staff to provide quicker and better support.

Leveraging vSphere, a trusted platform

One reason VMware has a great opportunity in the Edge space is that it can leverage its vSphere software stack, rather than building a whole new virtualization and management stack of its own. vSphere has been in the market for almost a decade, and has undergone a series of iterative improvements catering to customer needs in that time. Bugs have been fixed, performance issues have been addressed, and features have been prioritized after numerous customer interactions. vSphere has not only matured as a technology, but has also won the precious trust of thousands of customers along the way, making it a great candidate for early adopters of Edge computing. It also means that VMware can leverage the security, isolation, and performance primitives that are already baked into the vSphere platform, making go-to-market strategy a lot simpler.

Competition from other big players

Microsoft Azure

Venkat Yalla, a PM at Microsoft Azure, talks about their edge platform architecture in this awesome episode of Software Engineering Daily. Their idea is that it is reasonable to assume edge devices would comprise of server-class or, at the very least, desktop-class machines running linux. He describes an architecture that leverages Kubernetes to orchestrate deployment and management of edge nodes as containers on these machines. Their key customer-facing value propositions, in addition to the ones discussed above, are:

  1. Simple deployment interface
  2. K8s APIs for setup and management
  3. One-click deployment of code to the edge
  4. Automatic scheduling and scaling of nodes

An important internal-facing value proposition is that by leveraging Kubernetes and their own already existing virtual Kubelet for scheduling and other tasks, the development costs incurred reduce drastically, heavily influencing their go-to-market strategy and iterative abilities.

AWS

Through their predominantly server-less model, AWS Greengrass customers can set up their edge-computing environment and write Lambda functions that will run the business logic that they want to be executed at the edge nodes. While this comes with the guaranteed downside of vendor lock-in, the rich set of cloud services exposed to AWS public cloud users will be available to consume at the edge. Since the majority of edge applications are predicted to be using some sort of machine learning, AWS SageMaker is made available to consume directly by the code that is deployed to the edge.

Other important players, like Nutanix, are also making inroads of their own into this near-future vision of edge devices running workloads for machine learning based applications.

VMware’s differentiator: Cloud-agnostic approach

So what’s VMware’s biggest differentiator? If Project Dimension is executed well, and is available on multiple public cloud IaaS provider offerings, their cloud provider-agnostic approach may set them apart from the rest. The threat of vendor lock-in continues to be one that the market is weary of, and removing that pain-point may prove to be a huge factor in the way enterprises adopt this new technology.

Leave a Reply