Why vSphere users should migrate to open source cloud today
Enhanced with new formulas for managed services, “Open Source All the Way Down” is now a practical alternative to proprietary cloud architectures for popular enterprise use cases – both in datacenters and in public clouds
Continuing drama over Broadcom’s proposed acquisition of VMware has users still scratching their heads over whether the deal will proceed. The larger narrative, though, never questions VMware as a best-of-breed technical good – though perhaps one whose business-level utility to customers may be eroded by the company changing hands.
Why not ask that question? Is VMware – regardless of ownership – still tech-and-business justifiable as the default choice for enterprise private cloud? Clearly, a lot has changed the past few years:
Open source has caught up. VMware is great. But nowadays, solutions based in open source and running on generic hardware (or public clouds) can mostly match VMware and other proprietary cloud frameworks (for example, public clouds, which are also, in effect, proprietary), feature for feature and characteristic for characteristic – with much lower TCO, lower operating costs and staff cognitive drain. And of course, without lock-in: which always comes with costs, and limits freedom to innovate.
Cloud frameworks aren’t the focus any more. IT organizations have figured out that what matters more than technology brands is upsides: faster software delivery, more highly-automated Dev and Ops – gained from being able to focus attention on business imperatives instead of core platforms and infrastructure. The latter must be fungible: lower-level infrastructure “stacks” should be abstracted, so adapting to new ones (or dealing with multiple infrastructures) is easy, and higher-order logic (for example, application level automation) can be consistent. Otherwise, useful paradigms like hybrid cloud (cloud on your premises, cloud on a public cloud provider, both functioning in harmony) become much harder to implement.
Platforms are where the action is. Whether or not they’re aware of it – what matters to enterprises today is “platforms,” in the new meaning of the term. Five years ago, a “platform” was a cloud framework, like VMware vSphere or Openstack or AWS. Today, a platform is an integrated system of software, automation, engineering and operations expertise and services for building and running applications in production. Ideally, platforms should be commodities: you should be able to mount your platform on any infrastructure and have it look the same to applications, developers, DevOps.
The goal of a platform is to liberate technical talent to focus on delivering value. Modern platforms are supposed to do this, without creating lock-in and/or preventing rapid adoption of new technologies, like containers – tech that most technologists already acknowledge is becoming “how software is built and run.”
Diminishing returns on legacy models
What do organizations value about VMware? There’s a lot to like, as we said above. But the architecture of vSphere and vCenter also represent a snapshot of a point in the past, based on old assumptions about how hardware and data centers need to work to be efficient and reliable. Meanwhile, recent technical trends and business drivers have, in many ways, changed the game.
Users most often cite two main reasons for sticking with VMware:
The extraordinary historic compatibility and stability of the ESXi hypervisor
Key automation features (such as vMotion, vSphere HA, Distributed Resource Scheduler), enabling node-to-node workload migration, high availability, and VM orchestration
Let’s look at these more closely.
ESXi - best of breed for a declining niche
Simply put: the VMware foundational hypervisor can typically host (that is, as a virtual machine) any suite of software that runs on any even vaguely contemporary guest operating system (Windows or Linux) on any kind of vaguely contemporary hardware.
It accomplishes this by taking over an entire host machine (of particular compatible types) and incorporating all the functionality of a host operating system (in other words, it is a “type 1” hypervisor). To some folks, this sounds great. You lock down the host environment by eliminating all the cruft of a complete operating system, reducing potential attack surface.
But at the same time, it’s limiting. Operating system vendors add a lot of value to modern operations, such as the ability to manage bare metal hosts en masse, across an entire enterprise. VMware and its certified resellers provide solutions for host management, sure, but this stuff costs and is out of scope with other solutions you’re probably using, making things inconsistent.
And it’s not as secure as you might think. As of February 8, an exploit against unpatched ESXi hosts, called “ESXiArgs,” has been used to execute a wide-ranging series of ransomware attacks with at least 2,803 victims, and potentially thousands more. In part, this vulnerability exists in the wild because VMware is hard to update.
What’s harder to see, though, is that the above “advantages” are no longer hugely relevant to enterprises that aren’t still heavily leveraging their unique characteristics. Meanwhile:
Workloads are not as weird as they used to be. Sure, some legacy “pet” workloads still exist. Some business models are locked into using monolithic, closed-source software distributed on VM images, and so forth. But modern applications are increasingly built in modular fashion, and are increasingly deployed on container environments, limiting dependency issues. It’s well past time to start thinking of hardware (and host operating systems) as commodities.
KVM is a type 1 hypervisor, built into the Linux kernel, and host Linux+KVM operating systems can be stripped way back to reduce attack surface. Technically, at this point, a Linux host supporting virtual machines on KVM can be configured to be just as performant and reliable as an ESXi host, and easier to update – helping ensure that CVEs get patched.
As you can see, open source solutions aren’t the only things to have evolved. Software development and operations best practices are evolving too, and the future is definitely not about the huge, heavy, demanding monolithic software architectures VMware evolved to support in its heyday.
Automation - doesn’t Kubernetes do that?
VMware automation has always been impressive. When you control the entire host software stack, you can do amazing things like taking fast snapshots and migrating whole VMs from one chunk of hardware to another. In the past, these features helped organizations avoid real crises, save real time, and avoid millions of dollars in downsides from application non-availability – often caused by human error.
But the basics for this kind of automation have existed for years on OpenStack – the open source cloud framework. And more recent engineering advances have very successfully married OpenStack to Kubernetes as a substrate, creating a VM-hosting environment that can be updated with minimal disruption, and that self-heals around control plane hardware failures. In the meantime, rapid evolution of application development towards containers and cloud native design patterns is working to make, for example, restart of application components downed by hardware failures into a basic, expected platform behavior instead of a special feature.
Wither legacy workloads?
While it’s clear that most organizations need to maintain the ability to host legacy and essential VM workloads – it makes sense to start questioning how much this should cost. In most cases, you can assume the number of such workloads, and their requirements, will be diminishing over time, and that critical workloads will gradually be modernized, making it ever more important to view VM-hosting capability as a target for potential cost reduction. Continuing to host an ever-smaller number of monoliths on premium-cost, proprietary private cloud makes less and less sense.
Is IaaS mostly a tool for hosting container clusters?
Increasingly, yes. Sure, critical legacy applications still need to live in VMs if you don’t modernize them. But open source IaaS is clearly taking on a commodified role. In fact, so is Kubernetes: in a truly modern cloud environment, the “platform” (see definition above) starts with Kubernetes, but then radiates outward to host developer workflow automation alongside operationalization frameworks like Lagoon, which let developers “just push their code” and stand back as Lagoon and Kubernetes conspire together to operate it in production, reliably, observably, and securely.
Meanwhile, critical operations functions for open source-based IaaS, Kubernetes, and software development/DevOps can be outsourced with stunning efficiency, at scale – enabling organizations to focus on what’s important for the business.
The changed role of IaaS and the changing needs of organizations are what really provide the impetus to seriously consider moving forward towards a VMware-free future. If not to retire VMware (nobody will do this), then to ensure that it occupies a limited footprint in an increasingly open-source-based enterprise IT estate.
Attend the webinar
On March 9, 2023 at 12:00 PM PT / 3:00 PM ET, Mirantis’ Daniel Virassamy, Principal Solution Architect, will deliver a webinar titled Why vSphere Users Should Migrate to Open Source Private Cloud Today, in which he’ll discuss (and then, a few days later in a smaller setting, demonstrate) this new approach to open-source-based enterprise cloud. He’ll deliver a lot of pragmatic value for anyone considering whether or not VMware should continue to dominate their organization’s technological roadmap. He’ll discuss real alternatives, and:
Articulate how they stack up against VMware
Show how they improve on the legacy model of IaaS by addressing real software development and enterprise IT needs
Outline how new solutions – combining hardened open source with expertise and automation – are implemented
Talk about what migration involves - and why it might be easier than you think
Please join us!