At VMworld, various cool new technologies were previewed. In this series of articles, I will write about some of those previewed technologies. Unfortunately, I can’t cover them all as there are simply too many. This article is about ESXi on ARM, which was session OCTO2944BU. For those who want to see the session, you can find it here. This session was presented by Andrei Warkentin and Daniel Beveridge. Please note that this is a summary of a session which is discussing a tech preview, these features may never be released, and this preview does not represent a commitment of any kind, and this feature (or it’s functionality) is subject to change. Now let’s dive into it, what is VMware doing with ARM?
First of all, what caught my interest in this session was the fact that Hivecell was mentioned. Hivecell is a rather unique solution which allows you to stack ARM hosts. What is unique about that? Well when I say stack, I mean stack in the physical sense. The interesting part here is that only the first node will need to have power and networking cables, and the rest will receive power and networking through a magnetic link. William wrote about it extensively, so go here to read more about these guys. Really cool solution if you ask me.
The session started with an intro to ARM and the various use cases. I wrote about that extensively when Chris Wolf and Daniel discussed that at VMworld 2018. So I am not going to reiterate that either, just click the link to figure out why ARM could be interesting. What was new in this session then compared to last year? Well they showed a couple of things which I have not seen shown in public before.
First thing that was discussed was the fact that VMware is looking to support the AWS ARM instances (A1 instance) which were introduced a while ago. The plan is to not only support ARM, but also support Elastic Network Interfaces(ENI) and Elastic Block Storage(EBS). All of it managed through vCenter Server of course. VMware is now looking for early validation customers and partners.
The second thing that was revealed was the fact that vSAN is now also running on an ARM-based cluster! Andrei and Daniel showed it working in the AWS Datacenters. Last year this was already briefly discussed when the “vSAN Witness” as a use case was mentioned. This year Andrei validates that that is entirely possible! We are not(!!) talking about using the Raspberry Pi as a witness, this was a cool gimmick for the keynote last year, but doesn’t have sufficient resources today to be used as a witness. The MacchiatoBin, however, can be used as a witness, and this is a tiny device. What is interesting here is that they validated a combination of x86 hosts for vSAN capacity and an ARM-based vSAN witness to run, and it did!
What was also discussed during the session was running ESXi on a SmartNIC. This was also shown during the Tuesday morning keynote by Greg Lavender (VMware’s new CTO). Very interesting solution which now allows you to run multiple hypervisors on a single physical box. In the example Daniel and Andrei had, they showed 4 hypervisors on a single physical box by leveraging 3 SmartNICs, and they also mentioned that they have this working for various SmartNIC vendors and flavors. Why would you do this? Well you could, for instance, have specific workloads that were designed and optimized for Arm running on the SmartNic while the server is an x86 box and is running regular traditional workloads.
I think it was a very interesting session, with some good reveals, and I am already looking forward to what these guys will have in store for us at VMworld 2020! Anyway, I can’t wait to start testing an Arm based vSAN Witness node. Great job guys.
‘The interesting part here is that only the first node will need to have power and networking cables, and the rest will receive power and networking through a magnetic link.’
So that if you disconnect the first node you screw the entire cluster, interesting indeed 😀
Andrei Warkentin (Arm @ VMware) says
They have battery and wifi. They can go full autonomous for an hour.
Yeah I asked the guys this question at the booth. Very interesting that they can switch to wifi during failure.
Or if you need to take anything but the top one down for any repairs, you have to loose more. I hope the design can manage that more than 1 of those have power and network .)
It would have be nice if they managed to create a loop, something like are usually connected storage enclosures. As it is, is one of the dumber implementations ever :-)))
Don Woodward says
I saw ESXi on an ARM system at a demo at VMWORLD 2018 in the Office of CTO both.
Andrei Warkentin (Arm @ VMware) says
The demos we showed in the Arm booth at VMworld 2019 Barcelona were waaay sexier. Actually – mindblowing!