Is VMWare Still Relevant in the Cloud Age?
Virtualisation and VMWare were interchangeable terms a decade ago, pretty much like tablet and iPad, or electric car and Tesla, are today. But the cloud has changed the rules of the game and the data centre, barring their necessity in exceptional industries, is dead. However, no one goes down without a fight. VMWare has rebranded itself as a “cloud company” and is actively throwing around buzzwords such as Hyperconverged infrastructure (HCI), hybrid cloud, multi-cloud, and Tanzu Kubernetes. Is there any substance underneath VMWare’s new clothes? Let’s find out.
Hyperconverged Infrastructure (HCI)
Let’s get started with VMWare’s flagship buzzword. I have never paid attention to HCI until recently, because I assumed that I knew what it meant. It is what it says on the tint, ain’t it? If you thought that HCI was some mega orgy of clouds and data centres, all becoming—converging—onto one single thing, you made the same—wrong—assumption I did.
HCI, instead, establishes VMWare’s vision that “all infrastructure should be commodity x86 hardware”, but isn’t this the case already? Almost. Remember storage arrays/Storage Area Networks (SANs)? Well, VMWare wants to replace those with software running on commodity x86 hardware. This software is called vSAN. In VMWare’s sales nirvana, infrastructure becomes hyperconverged when it is all based on commodity hardware and customers license vSphere for compute virtualisation, NSX for network virtualisation, and, drum roll, …, vSAN for storage virtualisation.
Is there any cloud relevance to HCI? The short answer is no. This is VMWare’s data centre game, as per their overarching vision of a “Software Defined Data Centre” or SDDC for short. vSAN plugs into AWS EBS though, and do you know what it is called when it does? Elastic vSAN! Brilliant.
Sarcasm apart, and being fair, VMWare is not really pushing HCI as a cloud alternative; they are more focused on fighting the storage virtualization battle with their nightmarish competitor, Nutanix.
Hybrid-cloud and Multi-cloud
VMWare game is about commotidising the hardware and adding value (i.e. selling a licence) via software. No prizes for those who guessed how AWS, Azure, and GCP are treated. Exactly, like undifferentiated infrastructure running under vCenter. It is also no surprise that VMWare “Cloud” is underpinned by the same usual suspects (vSphere, NSX, and vSAN) found in the data centre.
Sounds all good. Why be mono-cloud when you could be multi-cloud? Why be on-prem and have a cloud when you could have a hybrid cloud? Nothing wrong here, except that the word cloud is a misnomer. Let’s replace the word cloud with data centre. Hybrid data centre, multi data centre. That’s what VMWare really means. Naturally, though, a remote data centre is not a cloud.
See, VMWare’s vision is bottom-up, not top-down. Whilst VMWare’s slogan, “Any Cloud, Any Application, Any Device” suggests otherwise, their main staple is still the virtualisation of x86 hardware, regardless of whether it is found on-prem, or whether it is managed by Amazon, or Microsoft.
VMWare charges by node; not only they don’t care what their customers run; the larger and the more convoluted the customers' workloads are, the happier they are. For instance, consider a cluster of two hundred Microsoft SQL Server VMs: music to VMWare’s ears. Which is exactly what the cloud is not about. Remember what the W and S in AWS stand for? Yes, Web Services, not lame VMs and block storage. We want Amazon RDS rather than two hundred VMs. But HCI and hybrid/multi-cloud are just the bad records in VMWare’s sales jukebox. There is a complete, different, and brighter side to VMWare.
A Prelude to Tanzu
No one wants to be the next Kodak. VMWare knows very well that their traditional business model is increasingly one of musical chairs. In today’s digital age, the value is closer to customer experiences, not far down in the supply chain. If VMWare has commotidised the x86 platform, the cloud has commotidised virtualisation. Running a data centre-era hypervisor on AWS? That is surely not the future.
If you can’t beat them, join them. If you can’t join them, buy them. The acquisition of Pivotal (among others) conferred VMWare an application-centric flavour. So, in principle, VMWare cares about the apps, after all. Well, all I can say is the last thing developers want is to have Spring Boot tightly coupled with any clunky element of the legacy VMWare stack. But Pivotal is not just the Spring framework. There is Pivotal Cloud Foundry (PCF) and there is also Pivotal Container Service (PKS). But these two are problem children. PCF is a legacy, and PKS only helps VMWare join the ranks of the Red Hats and Ranchers of this world. There is no differentiation here; just more bloatware. This is why VMWare had to do something completely different and unique.
Tanzu, at last.
Tanzu, as everything in VMWare’s world, is a suite, so let’s focus on what matters about Tanzu which is its Kubernetes capability, better known as “Project Pacific”.
Project Pacific is about embedding Kubernetes seamlessly into vSphere leading to the convergence of containers and VMs onto a single platform. How does this fit with PKS? It doesn’t. PKS is the same as Red Hat’s offering, another layer of bloatware, on top of a hypervisor. Now, having the ability to orchestrate containers natively, without the wedding cake’s layers of pain, that’s what I call a game changer.
This is the thing; unless we are talking about legacies or databases, the “apps” on the server world are today containers, and the standard way to run them, scale them, and manage them is through Kubernetes. VMWare, when it comes to Tanzu, got the message.
I was sceptical about Tanzu (as in Project Pacific) in the beginning. I asked myself, isn’t this just dashboard-wise integration, similarly to what Oracle did with most of their acquisitions? No. There are no pigs and there is no lipstick.
VMWare is turning the container orchestration game on its head by transforming vSphere into a Kubernetes-first, VM-second platform. But wait a moment; there must be a fault in my narrative—you may say—because there is clearly a chicken and egg problem here. How can the vSphere be Kubernetes-first when Kubernetes itself has to run on a VM?
This is the thing about Tanzu, and why I am convinced that it is a game changer. VMWare is proposing the following vision: VMs should be treated like regular Kubernetes resources. What? Remember BOSH? Enter the Kubernetes Cluster API. Whilst there is a minimum element of bootstrapping to solve the chicken and egg paradox, the vision is that Kubernetes should be the uber supervisor for both containers and conventional VMs. How cool is that?
What is more, Tanzu Mission Control (yes, there is a single pane of glass thingy!) can manage third party Kubernetes clusters, such as Amazon EKS, AKS, and GKE. This way, there is no need to roll out bloatware on top of public clouds, as per VMWare’s legacy hybrid and multi-cloud myopic sales agenda.
Render to Caesar the things that are Caesar’s; and to God the things that are God’s. VMWare’s strong suit is the data centre. Running their stack on a public cloud is fine for on-prem first enterprises (especially for disaster recovery purposes) but a terrible idea for any organisation with an agile, digital outlook, barring, as said in the beginning, exceptional use cases.
VMWare was losing the plot trying to dress up their old virtualisation stack as something valuable in the cloud. PKS wasn’t the answer either because there are managed Kubernetes services in the public cloud. The value in Kubernetes lies in its interoperability as made manifest by the wealth of managed platforms that abstract away Kubernetes itself and whatever infrastructure, virtualised or not, sits behind it. If companies have to “standardise” on PKS, Red Hat OpenShift or Google Anthos everywhere, then something has gone wrong.
Why is Tanzu a good thing then? Well, by definition, the data centre has to be managed by the owner (or their contractors), so here we want an experience that is the closest to a managed equivalent in the public cloud. And this is exactly what Tanzu is meant to be; the most seamless Kubernetes experience on-prem. Of course, VMWare would suggest that you could run Tanzu in the public cloud as well. Pssst….don’t buy into that part of their pitch.
Similarly to Microsoft, Oracle, Apple, and others, VMWare has both good and bad products. Actually, it’s not that VMWare has some terrible products per se, when it comes to their HCI and hybrid/multi-cloud non-strategy; they are merely part of an old fashion, data centre-era offering, that has been dressed up as something else. But Tanzu, on the other hand, is exactly what the market was waiting for.
What about their competitors? Whilst the likes of Microsoft Azure Arc, and Google Anthos bring the cloud on-prem, rather than the other way around, they haven’t cracked the chicken and egg problem of how to bootstrap the platform in the first place in the seamless way that Tanzu does via its beautiful Kubernetes Cluster API approach—Amazon gets that software is a hassle and this is why the likes of AWS Outposts and AWS Snowball are self-contained appliances.
Google Anthos, in particular, whilst managing the process of standing up (and upgrading) Kubernetes nodes' VMs, still relies on an underlying virtualisation platform. Guess whom?
For on-prem first organisations who already run VMWare, Tanzu is the logical upgrade path and a sound approach for migrating VMs into containers, in a gentle, organic fashion. It also opens the door for less traumatic migration strategies other than rushed and chaotic lift and shifts. For example, organisations can rationalise their existing data centres using Tanzu, by moving existing workloads to containers first, and at a later, convenient time, move said containerised workloads to a managed Kubernetes service in the public cloud.