How OpenFaaS Found Its Groove with Arm
After spending years at the cutting edge of both cloud…
For the last 20 years, VMware has helped make staying connected a more manageable task for countless organizations in nearly every industry imaginable — from Fortune 500 companies and universities to the Academy of Motion Picture Arts & Sciences.
The Silicon Valley software titan pioneered x86 server virtualization and its customers across the globe have benefited from reduced operating costs, streamlined configurations, rapid deployment, and better visibility for their IT infrastructure.
In the last few years, however, the surge of IoT and Edge use cases has generated a dynamic new class of workload — one that makes the way data is processed and delivered exponentially more challenging not only for consumers, but also for the companies that depend upon VMware. Thinking outside the classic datacenter has become a technological imperative.
The prevalence of Arm-based devices in the Edge ecosystem hasn’t escaped VMware’s attention — and a “technology exploration” between the two companies continues to bear fruit.
“It is clear that Arm has a lot of mindshare at the Edge,” says Daniel Beveridge, Director for VMware’s Advanced Technology Group in VMware’s office of the CTO and self-professed “cheerleader” for embracing a multi-platform posture, making Arm “an equal citizen in our management tools, and giving customers a unified experience from cloud all the way to Edge.”
At its annual VMworld event in 2018, the company kicked the covers off their plans to explore opportunities for ARM-powered use cases at the Edge. While plainly stating it had “no plans to support Arm in the datacenter at this time,” VMware demonstrated their ESXi hypervisor running on a 64-bit Arm device that was part of a wind turbine farm.
For VMware’s CTO office, the project was about exploring how to develop and deploy services for Edge use cases that are defined by smaller form factors, limited resources, or a variety of hardware from different OEMs.
The same capabilities that help customers run workloads confidently in production or upgrade systems without disrupting applications offer value at the Edge — particularly for use cases where there’s a lot of money or overall risk at stake.
“When you get to what we call the far Edge, where there is limited reliability and resources, the technologies we have in the platform such as High Availability, Fault Tolerance, or vSAN, bring a lot of value” says Beveridge. “And the reason is simple: You don’t want a single point of failure, such as a small storage device, to bring down your application or your whole system. As just one example, our software allows you to have a shared storage system and host to host memory replication so that any node can go down without taking down the application.”
According to the CTO team at VMware, the collaboration with Arm aligns with their long-standing goal of helping customers run workloads efficiently and confidently, no matter where they are or what hardware is leveraged.
VMware’s work on Arm is important for bringing this same experience to the Edge, says VMware’s Arm enablement architect Andrei Warkentin. The company is examining several use cases and resource requirements at the Edge:
For IoT users, VMware offers key capabilities such as workload and application lifecycle management at scale.
“Virtualization is not a silver bullet, but some of the problems that one has to solve for a successful IoT Edge deployment we have already solved for customers in datacenters and in the cloud,” says Warkentin. “Edge computing is inherently about solving IT and OT challenges at a massive scale. It is about treating the entire planet as a datacenter, with a cloud-like operational model.”
The workloads that VMware’s Arm team is targeting are greenfield in nature as more businesses develop customized stacks for remote locations. And at the Edge, since there are likely multiple systems running, ensuring connectivity is both a challenge and a means of protecting against data loss.
One of the key differentiators of ESXi for Edge Compute is the ability to reliably run mission-critical workloads non-stop, using VMware Fault Tolerance technology.
“For example, if you’re operating and monitoring an offshore oil rig on the far Edge, you can’t afford for the control and monitoring system to go down because of the high cost of on-site repairs and potential operational losses,” says Warkentin. “What Fault Tolerance does is run a critical workload in lockstep across two Edge gateways clustered together. So if one of the systems catches on fire and breaks, the workload transparently continues operation on the other gateway. This also allows for seamless maintenance while keeping all applications running.”
While Edge virtualization is an area of investigation for VMware’s work with Arm, “there’s also infrastructure work around Arm that we’re doing,” says VMware Group Product Line Manager Will Pien.
With more infrastructure moving into the public cloud, the VMware team had also been working with AWS on a version of their hypervisor that runs on AWS’ Arm-based Graviton and Graviton 2 silicon, Pien says, adding “By enabling our hypervisor to run on the Graviton, customers can potentially create a hybrid architecture environment where they run some applications on the Graviton, and others on the x86 in a performance/cost optimized mix.”
As VMware’s Edge strategy continues to evolve and the team explores new productization opportunities, they’ve made a few pivots along the way —including ESXi running directly on the SmartNIC.
In its basic form, a SmartNIC is an accelerator for network traffic. But SmartNICs also have capabilities that go beyond processing packets. With dedicated CPU as well as RAM, SmartNICs can operate as a small, dedicated computer, offloading tasks that would normally bog down a server’s main CPU.
“Telcos are interested in this use case because they can do some processing on the NIC without going to the main processor,” says Pien. This is not only more cost efficient, but can also dramatically improve performance for the network-focused workloads driving adoption at the Edge.
Though there are currently no plans to release ESXi on Arm as a supported product, the VMware team is considering plans to make a trial available to a broader audience, along with installation guides and recommended hardware platforms in small, medium, and large flavors.
“And those can range from a datacenter class of server down to a Raspberry Pi,” says Beveridge. “So we’re going to cater to the Enterprise tire kickers as well as pro-sumer-type geeks who want to check it out and we think we’ll learn quite a bit about what users are able to do with the technology. That will help us make future decisions on the direction that we want to take.”
From an engineering perspective, says Warkentin, “ESXi on Arm has been well-received simply because it allows us to evaluate our hypervisor through the prism of different hardware. We looked at the code from a new perspective and arrived at a better design. So working on Arm has made our hypervisor a better system on x86. That’s a very surprising outcome — and a very clear one.”