VMware Edge Compute Stack at the Industrial Edge
VMware presented its vision for edge computing with an updated Edge Compute Stack (ECS) offering to delegates at the Edge Field Day 3 event.
But first, we needed to forget what we know about VMware, a Broadcom company, aided by a gadget from the “Men in Black” movie: the neuralyzer. (That’s the device that wipes the memory of a target.) It was a funny way for Alan Renouf, technology product manager, and Chris Taylor, senior product marketing manager, to highlight that ECS is part of VMware’s Software Defined Edge (SDE) Division.?
What is ECS and VECO
The backstory to the joke is that there's a significant amount of backlash with changes to VMware Cloud Foundation (VCF) licensing and support costs. However, Renouf pointed out that VMware Edge Compute Stack (ECS) is not synonymous with VCS.
“We have reimagined what we've done at VMware to make sure that it fits the constraints that we see at the edge,” according to Renouf. There is also a distinct pricing and GTM for ECS that is appropriate for edge environments, he noted.?
ECS accounts for unique challenges that aren’t present in the enterprise data center or cloud environments. Among the challenges:
- Scale (managing applications on thousands (or tens of thousands) of devices.?
- Connectivity (applications that need to be able to run disconnected from the network).
- Manageability (there are few or no onsite personnel to maintain devices).
The ESXi hypervisor is a core component within the VCF architecture and relevant in ECS as well. In ECS, the edge runtime uses an optimized version of ESXi to run VMs and can run Kubernetes containers as well via Tanzu Kubernetes Grid. This provides flexibility for running a wider array of applications than other edge environments.
The other component of VMware’s edge offerings is Edge Cloud Orchestrator (VECO), a SaaS-based orchestration tool that enables zero-touch provisioning of edge hardware and applications. With a major update announced earlier this year, the company implemented a “pull” architecture, meaning an edge agent uses a heartbeat/keepalive signal to authenticate and signal state to the management plane to deal with the larger scale of edge deployments. Other new features include observability for infrastructure, VMs and Kubernetes workloads based on included implementations of Prometheus and Grafana.
Customer reference — Audi
The EFD3 presentation covered the architecture of VMware’s edge solution, and its use in manufacturing, with Audi mentioned as a reference customer.
We dug in after the event and found Audi has detailed its use of ECS as part of its “Edge Cloud for Production” (EC4P) initiative at the B?llinger H?fe plant in Germany.?
One goal for EC4P is to consolidate industrial PCs (IPCs) and programmable logic controllers (PLCs) on the shop floor. For example, a plant usually has separate IPCs for different functions, including the Manufacturing Execution System (MES), computer vision (for quality control), tool control, and predictive maintenance.
This approach is costly in terms of energy consumption, troubleshooting, and maintenance — the latter of which could also lead to security issues because of the complexity of patching different systems.?
领英推荐
Audi is using VMware to consolidate workloads by using a single platform for virtualization and containers. The goal: lowering capital and operating costs while gaining the ability to quickly integrate new tools.?
The workloads can run on a local cluster but also leverage nearby edge clouds where appropriate. This leads to the next point: what is unique is that both real-time and non-real-time workloads are supported.
Both VDI and VMs can run on the edge hardware using ESXi; Audi has the MES running as a VDI connected to a nearby edge cloud that is anywhere from 200 meters to 10km away, for example.?
Meanwhile, virtualized PLCs take data from sensors and actuators on the vehicle production line and then control the production robots. The vPLC systems need to execute commands within a span of five to 10 microseconds, according to Audi, with a total performance budget of 250 microseconds. This compares to the VDI workloads, where there is a goal of 25ms latency for non-real-time (RT) workloads.?
Other items of interest:
- A single vPLC can control a robot “cell”, which is a group of 8 robots; executives said vPLC instances have been running since the beginning of the year without interruption
- The company is running virtual PLCs (vPLC) on ESXi RT because it is a known technology, Audi executives said.
- Audi expects that VMs will be running in plants for the foreseeable future because there are too many workloads that have dependencies on Windows.?
- In the long term, perhaps by 2027, Audi expects a more even distribution between VM and container workloads as more net new applications are created for Kubernetes.
Key Takeaway
Audi’s use of VMware is a great example of edge computing at scale for both server room and shop floor workloads. According to Audi executives, a typical factory has over 10,000 industrial PCs and 4000 human-machine interface systems used in the control and monitoring of production.?
In the year-plus since entering the production phase, Audi has been able to reduce the number of dedicated function devices while improving IT operations with a single operating model and processes for managing workloads at the factory.?
Analysis
VMware’s ECS addresses the needs of its target customers with the ability to run and manage VMs and containers alike. Those customers are likely to be large VMware shops that want to incorporate edge applications into their broader cloud infrastructure.
Will it be useful beyond that market, and is that even a consideration??
Enterprises are unlikely to forget what they know about Broadcom. There is negative sentiment about VMware that’s impacting spending plans at some organizations (link at the bottom of this page). Meanwhile, the company has reportedly decided to sell directly to its top 2000 customers.?
There is plenty of opportunity for VMware to grow ECS at large enterprises, but we think smaller enterprises that have historically been served by channel partners are likely looking elsewhere. We just returned from a channel partner event where it was clear that competing cloud providers are spending big money to help shift customers away from VMware — one provider was waiving $1.5 million in fees for the first year on a $3 million/year deal. Meanwhile, one customer said they were paying an extra $900,000 for VCF licensing because it was going to take another year to get off VMware. It is hard to see those customers like that investing in VMware for new edge computing initiatives.