KVM, the default hypervisor in OpenStack cloud software, faces increasing competition from VMware ESX Server. Mirantis cofounder Boris Renski puts the race in context.
InformationWeek Editor’s note: Mirantis’ cofounder and EVP Boris Renski said on September 11, 2012, that the OpenStack Foundation’s accepting VMware was a mistake. Now he finds himself eating some of those words.
So far, open-source KVM has been enjoying uncontested leadership as the default hypervisor for OpenStack cloud software. That’s largely because OpenStack runs on Linux and KVM comes standard with all Linux distributions. A recent survey of 822 OpenStack users by the OpenStack User Committee clearly illustrates KVM’s predominance. KVM’s share is represented by 86 users, versus 6 ESX users on server clusters with one to one hundred cores. On larger clusters of 101 to 500 cores, the gap remains large: 33 KVM to 4 ESX.
However, we are starting to see trends that indicate that KVM is being challenged by ESX Server. About 18 months ago in my presentation at the OpenStack Summit, I shared my vision of the OpenStack adoption curve over time. The year 2014 would be when enterprises start seriously looking at OpenStack. Today it seems like my prediction was correct. Looking at Mirantis’ customers and pipeline, we see a significant spike in uptake of OpenStack by traditional enterprises, particularly those in financial services.
When it comes to the enterprise market, we all know that VMware’s vSphere virtualization management system is king. While we can argue day and night whether the OpenStack/ESX Server combination makes sense, one thing is for sure: enterprises will not do away with their existing VMware investment overnight. This means that as OpenStack gains its enterprise footprint, it will be increasingly used to offer self-service capabilities on top of vSphere infrastructure, which in turn will bump up ESX’s share among hypervisors being used with OpenStack.
This year VMware made its long-awaited hybrid cloud service announcement. Good news for VMware. Bad news for cloud service providers that built their business on VMware technology. Now they have to compete with the vendor whose technology they’ve been reselling for years. VMware service providers now have to hedge, and OpenStack is the answer.
Again, throwing away VMware and retooling the entire stack is not the answer. Running OpenStack on top of vSphere translates into more customer choice and hedging against VMware over time. The result: more OpenStack with ESX.
VMware also announced the split of vCloud Director into vCloud Automation Center (vCAC), a product that’s derived from VMware’s DynamicOps acquisition. This left a gap in VMware’s product portfolio around DevOps-focused, self-service functionality. vCAC is heavyweight software, much like an IT service management product, requiring deep integration with IT business processes and an ERP-like implementation scenario. Pure vCenter, on the other hand, is built for system admins and lacks the cloud-like self-service features that WS has established as a standard for the next-generation engineering ecosystem.
Moreover, enterprises today are concerned about vendor lock-in. As they consider moving beyond virtualization toward a private cloud, they can choose to either invest more in VMware, or proceed on a tactically challenging but strategically rewarding path of open – as in open-source and OpenStack.
Running OpenStack on top of vSphere is a sensible choice that can edge enterprises closer to the cloud. On the one hand, they get the self-service functionality of discontinued vCloud Director. On the other, they take a step toward liberating their stack from vendor lock-in.
Last but not least, despite my earlier speculations, VMware actually did make a very tangible commitment to ensure that OpenStack runs well with VMware products — a sensible move if the company wants to stay relevant in the datacenter.
While all these factors are unlikely to dethrone KVM from the number one spot, they will make meaningful headway in favor of ESX in 2014.
I believe this is a positive trend. After all, there are two camps of thought on OpenStack. The first believes OpenStack is a bunch of raw open-source code bits meant for commercial vendors to reuse in an opinionated, integrated solution. The second believes it is an open, standards-based fabric for federating diverse pools of infrastructure. The first camp equates OpenStack to just another messy open-source project, the second to a highly disruptive, industry-changing phenomenon. At Mirantis we put our money on the second view. A healthier diversity at the OpenStack hypervisor layer will edge OpenStack closer to that vision.
This article was originally published by InformationWeek on December 18, 2013.1 comment