Last year, I wrote an article questioning whether Citrix or VMware could find genuine opportunity for revenue with IoT. Shortly afterward, Citrix dropped the Octoblu team, around which much of their IoT focus had been.
Meanwhile, the measured approach of VMware has continued with their Pulse product. And at VMworld 2018, we saw some interesting announcements, including ESXi for ARM x64, indicating VMware are taking quite seriously now.
At one time, Citrix seemed a far better fit to the embedded world than VMware—however, that all seems to have now changed.
Citrix and IoT
Open-source Xen running on ARM architecture was a natural synergy with the Citrix open-source , a location that is also home to ARM’s main development and research site. Since 2013, the Xen hypervisor community has been working on drafting and developing a project titled, “Embedded and Automotive PV Drivers/Project Proposal.”
There was a nice overview published in 2014, but little noise since. In 2017, there were some tangible developments and progress, with the Linux Foundation recognizing the need for a hypervisor in this space; it was speculated that Xen—alongside KVM, OpenVZ, and LXC—could be a contender.
The project aims to allow Xen to run on embedded devices and architectures such as ARM, rather than traditional more powerful server architectures. A hypervisor on can provide the isolation, security, and reliability needed to enable features such as trusted root and segregating drivers/applications from different vendors to avoid interoperability issues.
Citrix’s IoT strategy, though, has been rather muddled. Although they’ve had access to the raw Xen hypervisor technology to run on ARM architectures, this hasn’t found its way into commercially supported Citrix XenServer, let alone a wider EMM or IoT strategy for the mainstream revenue products. Citrix has lost some of the original architects and protagonists for ARM on Xen from their in-house open-source Xen team, notably Stefano Stabellini (now at Xilinx, Inc).
At one time, Citrix probably had the lead on IoT futures, ARM support, and certainly on their messaging; lots of blue sky demos especially from the acquired (then axed) Octoblu team (Gabe questioned the alignment with Citrix’s core direction at the time) but it all never seemed to find its way into many tangible enterprise products. lots of LEDs (including an IoT Christmas Sweater), and even a talking rubber fish. At least Workspace Hub finally got off the ground in 2018.
VMware, on the other hand, seem to have a rather clearer idea of where they wanted to be in the space, and now that includes ARM support for ESXi. Being a closed-source hypervisor, it’s always a lot harder to guess at what might be coming, let alone when. However, over three years ago (April 2015), an obscure whitepaper called “VMware Solutions for the Connected Car” (PDF) quietly appeared. It was a well-considered paper, indicating that someone at VMware was evaluating ESXi for automotive and embedded very seriously.
One paragraph that caught my eye was:
“VMware, the recognized leader in x86 virtualization, is now closely following developments in ARM's technology ecosystem and has a team dedicated to porting the VMware ESXi™ hypervisor to the ARM8 architecture. There is currently insufficient server market demand to justify releasing this as part of a product portfolio.”
Roll-on to VMworld 2018, and VMware announced a maturing Pulse 2.0, a product designed to include IoT gateways and device management in the datacenter infrastructure. About 98% of datacenters currently use Intel, but last year Microsoft announced their intentions to diversify to ARM, and new ARM-based servers from ATOS came to market. Hopefully, this will eventually lead to more competition (and a better value)
With ARM laptops starting to appear as well as servers, not to mention the trillions of embedded ARM chips; VMware seem to have quietly crept up on this market. Meanwhile, it’s still not really clear where Citrix is heading, especially on the hypervisor front. VMware, on the other hand, now seem well placed now on IoT gateway integration and ARM for servers (although at the moment they seem focused on just IoT edge compute cases).
We haven’t yet seen the hypervisors made by EUC vendors foray into the automotive/embedded world, but ESXi, with VMware’s solid enterprise history, c, which have already deployed VMware for their VDI infrastructure, including many Automotive and Manufacturing organizations. The potential market is certainly large, but it’s out of traditional EUC territory.
VMware’s closed-source approach also sits slightly ill at ease with the Linux ecosystem and culture of the embedded world and Linux Foundation. However, regulated enterprise tend to like commercially supported solutions—for example, when it comes to other applications like VDI, large automotive customers tend to choose ESXi rather than open-source KVM.
Whatever VMware do or don’t do with regard to automotive and embedded I suspect has been extremely well researched and evaluated. The details in the 2015 VMware whitepaper on how ESXi could find a multitude of uses in automotive are fascinating. It’s a great read on the technologies of the connected car for anyone; we can only wonder how those thoughts have progressed in the last three years.
A few more thoughts
A parallel announcement from VMworld involved Shawn Bass demonstrating three users (up to 10 is possible) sharing a session to work on the same CAD model. This is a feature the automotive industry have been crying out for from hypervisor vendors, with HPE RGS previously pretty much the only responsive option for collaborative screen sharing. This strengthens VMware’s place even further in automotive VDI and CAD/CAE application delivery.
Connected cars, transport, nuclear power, and utilities are already governed by Functional Safety (FuSa) legislation. The processes and investments needed to break into FuSa-driven IoT are stringent and demanding and it’s actually very hard to imagine an open-sourced ecosystem delivering those needs. Anyone fancy a VMware car?