Alex opened a large can of worms and the ensuing discussion and blowback has been riveting. But more important than the ensuing online chaos he created with his post are Alex’s blatantly incorrect assertions and assumptions.
In his post, Alex claims Mirantis customers prefer utilizing "a combination of Heat for orchestration, Trove for the database, LBaaS for elasticity, then glu[ing] it all together with scripts and Python code" rather than a traditional PaaS implementation, like Cloud Foundry. Alex believes his suggested approach provides a more "supported solution" than what’s available from PaaS vendors today. While it’s entirely possible that Mirantis customers haven’t yet experienced the benefits of PaaS, I believe his viewpoint ignores what is happening within the OpenStack community and its leading contributors. OpenStack will continue to develop services and those services will be helpful for PaaS consumers but I do not believe OpenStack will build a PaaS offering as a foundation supported project. This is especially true when one examines the outside activities of its leading contributors.
To really dive in, it is first important to establish a clear definition of Platform as a Service (PaaS). Salesforce, the parent company of Heroku, defines PaaS as “a proven model for running applications without the hassle of maintaining the hardware and software infrastructure at your company.” In short, PaaS was developed to ensure users don’t ever have to touch or glue together scripts.
PaaS, however, is about more than just speeding application delivery. In its most effective form, PaaS delivers application portability, service abstraction and developer satisfaction. PaaS should eliminate all server administration from day-to-day developer operations.
Heroku, arguably the grandfather of PaaS, was founded in 2007 and certainly had success attracting early adopters. Yet PaaS is just now emerging as a viable enterprise technology and is beginning to reach a level of acceptance with enterprise engineering teams. Remember, PaaS is not a replacement for IaaS. Rather it is a PaaS riding on top of IaaS that creates an incredibly powerful solution.
With that in mind, let’s come back to Alex’s assertion that OpenStack and Cloud Foundry are heading down a conflicting path. OpenStack’s Havana release brings us a series of new services, most notably, Heat and Trove. Heat is OpenStack’s equivalent of Cloud Formation and Trove is OpenStack’s equivalent to RDS. OpenStack is moving up the services stack, which is ideal for the OpenStack ecosystem, but it is still far away from developing a Platform as a Service from an application delivery perspective. Compute, networking and storage remain the underpinnings of any IaaS. With those systems now well established within OpenStack, it’s natural for the project to direct its efforts into supporting services such as LBaaS or DBaaS. But those supporting services do not replace the features delivered by a true PaaS. Claiming you can glue a few services together to replace all the features of a PaaS is ludicrous. If this were true, any IaaS provider could claim to be a PaaS provider. The PaaS market place wouldn’t see any adoption.
Furthermore, many of the top contributors to OpenStack’s Havana release are already working on a PaaS strategy. RedHat created OpenShift. NTT, HP (via ActiveState) and IBM all support Cloud Foundry. It’s highly unlikely these organizations will push a PaaS agenda within OpenStack as it works against their existing technology roadmaps and investments. Instead, I believe they will continue to push development of supporting services that will make a PaaS offering on top of OpenStack an even more powerful experience.
“Have you heard of the successful adoption of Cloud Foundry or OpenShift on top of OpenStack? Well, neither have I, and my prediction is that no such adoption is imminent.”
— Alex Freedland
A big portion of OpenStack’s original success derives from its ability to build an ecosystem and enterprise collective to support its core purpose. Cloud Foundry has done the same thing. As noted above, IBM and NTT, HP (via ActiveState) have successful Cloud Foundry implementations on public OpenStack clusters. AppFog, prior to their acquisition by Savvis, had a sizable Cloud Foundry implementation running on top of Rackspace’s public OpenStack Cloud. Blue Box receives inquiries every month asking for Cloud Foundry on OpenStack and already has Cloud Foundry + OpenStack implementations operating for customers.
Much like the OpenStack Foundation, Cloud Foundry has support from large, established enterprises and innovative technology startups - Pivotal, GE, IBM, NTT, Rakuten, Intel, Savvis and Piston. This ecosystem that will help establish CF as the dominant open source PaaS offering. Cloud Foundry has customers operating today at massive scale--Baidu is using Cloud Foundry in production to serve one billion page views per day. There’s a reason these companies are using this technology. PaaS is fundamentally different than IaaS and the evolution of additional services within OpenStack will not alter that fact.
Alex is correct that vendors delivering proprietary closed source software to the cloud marketplace should be concerned for their business model. These modern open source projects can evolve faster than any individual proprietary software can. But Cloud Foundry is as far away from “proprietary” as I can imagine.
PaaS is here to stay and I believe Cloud Foundry will live along side OpenStack and together they will create a compelling offering. I look forward to the release of Blue Box’s Cloud Foundry offerings – the market is ready. The future is bright and “The Future is Now.”
Have you read the news?
Rumor has it this cloud thing is taking off.