The PaaS market had a major false start in the period 2009 to 2012. The first PaaS vendors came to market with one of two premises
- “we’ve got a really great platform you can use it if you want to”. Good examples are Force.com (A PaaS derived from an IaaS – salesforce.com), Google App Engine and the original version of Azure
- “it’s a great place to run applications in a particular language” – good examples are Heroku (ruby) and PhpFog (PhP)
Since 2011 a second-generation of PaaS infrastructure has emerged which is exemplified by VMware’s CloudFoundry and Red Hat’s OpenShift. The biggest change between first and second generation PaaS is in the mindset. Instead of the first “P” in Platform as a service referring to “a Platform” it now refers to “any Platform”. In other words the job of the PaaS is to support any application in any language and deliver any set of services that any application might reasonably require. Whilst the first-generation PaaS was generally monolithic, the second-generation PaaS is usually capable of being implemented on a broad range of IaaS and/or virtual infrastructure, and the key factor is openness and diversity. Thus CloudFoundry can be implemented on OpenStack. The market transition was most clearly signalled by VMware when in August 2011 it abruptly cancelled its nascent VMforce PaaS – a collaboration with SalesForce.com to deliver a packaged version of Force.com. Instead VMware went forward with CloudFoundry, derived from a startup which had been acquired by its SpringSource subsidiary in 2009. Crucially CloudFoundry has been open sourced and a community of vendors has developed around it which is delivering a diverse range of features and variant products. Indeed VMware’s own version (CloudFoundry.com) currently looks lacking in maturity in comparison with third party variants. However that’s only because VMworld is coming up and VMware has been holding back the announcements.
So, what could happen at VMworld?
As I’ve mentioned in previous posts, a credible Open Source project needs both a Licence and a Foundation. The Foundation provides a governance model which allows other vendors to feel comfortable enough to consume the technology. You can do this two ways – either build your own Foundation (as Rackspace did for OpenStack), or engage through an existing foundation (as Citrix did with CloudStack and the Apache Foundation). CloudFoundry does not yet have a foundation.
VMware has clearly dominated at the virtual infrastructure level, and whilst CloudFoundry is likely a very small direct contributor to VMware’s current revenues, it could achieve equal market dominance at the PaaS level, as long as VMware doesn’t mess it up. There is thus a lot of potential upside and very little actual downside in doing something dramatic to reflect the reality that CloudFoundry isn’t in any way bound in to the rest of the VMware stack. Market speculation has focussed on corporate restructuring – perhaps CloudFoundry could be spun off as a separate entity. However, from an Open Source perspective there is another crucial element, you can’t have a credible Open Source project where the reference implementation requires closed source components. In other words CloudFoundry.com (built on a VMware stack) can’t drive the Open Source project. Whilst you can doubtless run CloudFoundry on a number of other IaaS stacks, particularly Amazon Web Services (AWS), it is clear that the most popular Open Source choice is… OpenStack.
We therefore suggest CloudFoundry should merge into OpenStack.
This has several major benefits
- The resulting foundation would retain a very clear market positioning and focus for technical and commercial collaboration
- OpenStack provides an existing governance model that can easily be adapted to include PaaS
- VMware (or EMC via CloudFoundry if we believe the spin-out rumours) would act as a counterbalance to RackSpace and help the dynamics of OpenStack
- CloudFoundry can leverage the existing engagement of major vendors with OpenStack
- The key Enterprise Players are coalescing around OpenStack (e.g. HP, IBM, Cisco),
- Citrix isn’t going towards OpenStack
- Microsoft… we’ll leave speculation about .NET and open source for another day.
- the existing members of OpenStack wouldn’t be requried to cough up more money to join a separate CloudFoundry foundation
- It would stop OpenStack doing something silly and building its own PaaS
- It would help drive RackSpace Cloud revenues and could reduce the drain of CloudFoundry onto Amazon AWS
- Increased Rackspace revenues are an incentive to Rackspace to let CloudFoundry into the party
- There’s no VMware software on AWS, and scant pickings for EMC
- It would stop Red Hat (who are sniffing around OpenStack) from doing the same thing with OpenShift – thereby gaining the vendor community that it currently conspicuously lacks
Of course it’s only speculation. If you read this after VMworld and I’m wrong, hey today’s headlines are tomorrow’s fish and chip wrappers, but if I’m right, you heard it here first!