VMware’s Cloud Foundry has been festering for the best part of a year now. It smells a little bit of lack of courage, and a lot of lack of focus. The body is still warm, but I fear EMC/VMware may have already snatched defeat from the jaws of victory.
I wrote a post a while ago about Why would a Developer choose VMware?, and the answer is that there’s a really decent set of technologies that you might choose to pick up on their own merits that happen to be owned by VMware. Cloud Foundry binds these (and other) components together into a platform to which you can directly deploy using some pretty good developer tooling embedded in Eclipse, for which there is a community of developers adding plug-ins.
It’s the plurality and dynamics of the open source tools, languages, and frameworks and the multiplicity of front-ends that are driving the developer marketplace. Every time I look at the components there is something new. Microsoft has its own problems with .NET. It’s too alien to fit in and it’s too expensive to compete, whereas VMware has some of the best pieces, much more of the fresh good stuff than, say, IBM, who really ought to be able to play this game better. But the plurality and dynamics are also the problem—the pieces come from various sources, and if they want to continue to own the pieces, VMware will have to continue acquiring software.
The pieces are open source, and so it becomes unclear who you are buying bits from, if anyone. When you speak to Red Hat about this stuff they seem very familiar with the beast they are dealing with. They are used to a Distro model, and they are essentially saying that it’s not just the ownership of the pieces that matters, it’s the fact that all the pieces come in a box and the box has the word Red Hat OpenShift on the lid, and if there are some pieces that are missing or don’t fit together, it’s up to Red Hat to fix it. You can have the box in two versions, public or private, but it’s the same box. In one case it’s pay-to-use. In the other case it’s pay-for-support.
When Red Hat start comparing OpenShift with Cloud Foundry, they tell you (perhaps slightly economically with the truth) that everyone using Cloud Foundry has forked the stream and there’s no standard version, and if you adopt any of these customized versions you’re into instant lock-in. And you can see VMware’s Cloud Foundry team desperately trying to consolidate their adopters around “Cloud Foundry Core“, the least common denominator version, and actually telling everyone that if they adopt ouside the core they risk lock-in, thereby reinforcing Red Hat’s message. It’s a textbook example of how not to market an open source community.
In the meantime, everything seems completely static at VMware with Cloud Foundry. It’s lost so much momentum over the last year. At VMworld we heard … nothing.
The problem here is that PaaS doesn’t really have much to do with virtualization.
Everyone seems to think PaaS goes on IaaS—I suspect long-term it doesn’t. I don’t even think you really need a hypervisor under PaaS. You probably need an O/S, but even that I’m not sure about. I read all these posts on detailed hypervisor features, and then I look at PaaS and think, “Why would you care?” This stuff gets addressed in the services and the middleware. The app server is distributed and replicated. The performance is driven by the caching strategy in the mid tier, which is distributed, and the backend database runs its own distributed replication (SQL or noSQL).
So this leaves Cloud Foundry and the other pieces as a great standalone set of technologies that doesn’t really need to leverage the rest of VMware. You can put them on vSphere, but why would you, and how would you compete on price with people who don’t?
It is in this context that Cloud Foundry finds itself as it transitions into a new VMware/EMC operating entity currently known as the Pivotal Initiative. This is a real parts-spin-special—a rag-bag of stuff that EMC and VMware have acquired, mixing middleware and application services, big data analytics, and a services team. The expectation is that this will become a separate company. But at least the parts bin doesn’t contain vSphere, so Cloud Foundry won’t be forced to pretend it will somehow drive up sales of high-end hypervisors.
However, the process won’t really go anywhere till next year (if at all). In the meantime, VMware is sitting on Cloud Foundry, which is is losing market share to… errm… Cloud Foundry. It’s in the bizarre situation where the plethora of companies that have adopted Cloud Foundry are ahead in the marketplace of the company that owns it. At some point I expect Cloud Foundry will have to buy back market share by acquiring its adopters, and the longer it leaves this process, the more expensive it will be. Or, someone may actually fork the stream (and there are precedents on a smaller scale, mainly involving Oracle, like Libre Office and Jenkins) and take Cloud Foundry leadership away from VMware.