Blog Post

Get Purchased or Perish: The Harsh Reality of Cloud Platforms

It’s getting hard out there for startups in the Platform-as-a-Service (PaaS) space that have to compete increasingly with large vendors, and today’s news that VMware is launching its Cloud Foundry PaaS offering (s vmw) just made life more difficult. It’s an uphill battle, because large software vendors realize they must embrace the future of application development in the cloud or risk becoming relics of the past, and many are choosing the former.

Names like Heroku, Makara and Cloud Foundry still exist in the PaaS world, but they’re now attached to (s crm), Red Hat (s rht) and VMware, respectively. Elsewhere, Microsoft (s msft) (Windows Azure), Amazon Web Services (s amzn) (Elastic Beanstalk) and HP (s hpq) (it claims) are all pushing their own PaaS efforts, and IBM (s ibm) and Oracle (s orcl) can’t be far behind. Depending on how you look at it, it’s either a really good time or a really bad time to be a PaaS startup.

Earlier this week, I had a conversation with Issac Roth, former Makara co-founder and CEO, and current director of cloud product marketing at Red Hat, who suggested that PaaS is pretty much an acquisition-or-perish market at this point. Roth didn’t always see the world this way, but his experience since Red Hat bought Makara in November has changed his worldview. Some people thought Makara would die inside of Red Hat, he explained, but his reality has been quite the opposite.

Roth is still driving Makara development within Red Hat, and the company is planning to release the details of its forthcoming PaaS offering at its user conference in May. Further, he noted, Red Hat “is putting a ton of resources behind this, and Red Hat isn’t nearly as big as Microsoft.” The suggestion, of course, is that Microsoft is putting even more money into its Windows Azure platform, a very real possibility given its Microsoft’s recent pledge to invest $8.6 billion on cloud computing R&D this year. Red Hat is operating at a billion-dollar run rate for 2011, and CEO Jim Whitehurst has explained to me on multiple occasions how seriously it takes its cloud computing business as a driver of future revenue. A serious monetary investment is crucial, he explained, because developing enterprise-grade PaaS offerings is hard work that requires serious resources of the amount that only large vendors can afford.

For example, although auto-scaling gets a lot of attention — and rightly so — that feature alone doesn’t make a PaaS offering. PaaS providers need to work on a variety of capabilities and features, Roth explained, including isolation, capacity management and scheduling, and if they don’t have a large vendor’s existing tools to leverage or the research funds to create their own, the only option is to integrate open source projects and hope they’re production-ready. It’s not cheap to run a public PaaS service, either, he added, when developers want low-cost product but PaaS providers still need to pay their IaaS providers at the month’s end. Red Hat can leverage its certified cloud partners; Microsoft, IBM, VMware and others can leverage their own data centers. “How can a little company differentiate and keep up?” he asked.

From what I can tell, Roth’s experience with Red Hat isn’t unique. Heroku CEO Byron Sebastian has told me that life has been great since his company became part of, and the latter’s reputation already has opened up new doors for Heroku in terms of big-time customers and partners. I think the Cloud Foundry team — which was acquired by SpringSource right before VMware bought SpringSource — would tell of a similar experience as part of VMware. The product certainly has received a much-needed facelift.

Homegrown PaaS appears to be doing just fine, too. Microsoft is wooing some very big customers to its Windows Azure platform — including, recently, Toyota (s tm) — and although AWS doesn’t disclose any numbers, it’s safe to assume its free Elastic Beanstalk service is doing just fine among Amazon EC2 users. It will be even better when AWS incorporates support for additional languages beyond Java, including Ruby support via a partnership with Engine Yard.

What this all means for PaaS startups like DotCloud, PHP Fog, CloudBees and others (although not RightScale, which actually might have a broad-enough business model and customer base to remain independent if it so pleases) is that if they’re not already fielding acquisition offers, they should expect to be doing so soon as the remaining software giants gets their PaaS acts together. If the phone never rings, it might be time for some self-reflection on whether the platform is viable in the face of all those deep-pocketed competitors.

Image courtesy of Ross.

4 Responses to “Get Purchased or Perish: The Harsh Reality of Cloud Platforms”

  1. I still say the PaaS market is probably a bit early to make a final call on who wins and loses. I tend to think we are still very early and there is a lot of room for the type of innovation that smaller companies can typically provide.

    There are also two potential benefits from today’s (very impressive) VMWare announcement. One is market awareness which can float all boats. The other is they can set a pricing umbrella much like IBM and BEA did in the app server market. VMWare charges about $2K/year for the vSphere, vCloud, tcServer stack on a CPU today and they will not want to undermine that revenue stream too much with their own cloud offering or squeeze their large service provider partners.

    Of course I have a bias with my involvement in two small Cloud vendors – CloudBees and eXo, as do Isaac and Byron have a bias to support their businesses.

    • Derrick Harris

      I agree that innovation isn’t dead and startups will continue to be important in driving it, but the core PaaS capabilities are fast becoming the domain of large vendors. Just like IaaS firmed up pretty quick with providers who had infrastructure and scale, PaaS seems to be firming up around software vendors with all the tools. Now it’s time to start looking for the next iteration of cloud computing.

      • Derrick, here’s my takeaway from your article:

        1) Heroku CEO agrees with recent Heroku acquisition

        2) Makara founder agrees with recent Makara acquisition

        3) Large vendors will out-innovate startups because they have more engineers

        4) Large vendors are leveraging their own datacenters to lock out startups

        None of this makes sense, for a few reasons:

        – The argument of price is a joke. PaaS is not about the cost of infrastructure. Sure, IBM and Red Hat’s “certified cloud partners” can sell cheaper compute that a startup. And Amazon can sell compute twice as cheap as anyone else. Who cares? Selling compute is not our business.

        – Although I agree that large vendors have privileged access to larger customers, I can tell you with 100% certainty that Makara and Heroku gained absolutely no technology or product advantage from being acquired. That somehow startups need “large vendor tools” to build a successful platform is simply bullshit. It’s the other way around.

        – As Bob pointed out, the PaaS market is extremely early. Salesforce, Red Hat and Microsoft are no doubt mighty competitors, but the overwhelming majority of developers have not yet chosen a platform. Startups move fast, they don’t face the dilemma of disrupting themselves and there are no competitive barriers holding them back. Everyone is free to innovate and reap the benefits.

        – Which brings me to the key point: this is all about developers. We are all in a race to make developers happier and more productive. Nothing else matters. That is how I will evaluate the success of VMWare, PHPFog, Azure, Red Hat and DotCloud. So far everyone has a lot to prove.

        • Derrick Harris


          I think what Issac is referring to with his comments about the benefits of large-vendor benefits has to do with being part of a trusted brand even among CIOs, and about incorporating the types of capabilities that large businesses still care very much about. In that case, integration with CRM apps, databases, scheduling software, security, etc., do become valuable, even if they’re throwbacks to legacy development.

          Regarding price of compute, his point, which I think is valid, is that although selling compute is not a PaaS provider’s business, it does present a source of overhead; PaaS prices can only go so low if providers are paying full price for the underlying compute every month. Microsoft et al aren’t buying into an IaaS provider’s margins when they sell services (despite the fact that they sometimes charge more in return).

          Still, as I mentioned to Bob, I don’t think PaaS innovation is dead at all, and of course much of that will come from startups. In fact, from what I’ve seen, I think DotCloud has a whole lot of promise and will do very well with developers. I just think having to attract them from under the shadow of VMware, Microsoft, Red Hat, etc., will make that job more difficult. And if you’re successful at it, some large vendor that wants in on that action will take notice.