Platform clouds can make enterprises all teeth and no tail

Platform No. 9

Pivotal will bring together Cloud Foundry, the vFabric virtualized platform cloud, which is a derivative of sorts of Cloud Foundry, as well as the Cetas Software and Greenplum parallel data warehouse and Hadoop products.

These things take time. It took a decade for VMware's ESX server virtualization hypervisor to become the dominant virtualization server maker in the data center on x86 iron, and it will probably take a decade for enterprises to absorb platform clouds.

"Platforms in general take a while to get adopted because it has to have some footprint or better still ubiquity, and then you need an ecosystem of developers and applications on top of it," says Jerry Chen, vice president of cloud and application services at VMware.

"That doesn't come overnight. It is a chicken-egg situation, and you get the platform out there and get people using it. The good thing about Cloud Foundry is that we are using Java and Ruby and other technologies that they are familiar with, so we are not asking developers to do anything new or unnatural."

Having a private PaaS version of a public platform service is not enough, obviously. It is not just about what side of the firewall you are on. Enterprise customers, distinct from start-ups or app development departments using platform clouds to create applications and test them quickly, need availability, measurability, and control features added to the PaaS. For production, they need to add compliance, security, visibility, and other features.

"There's a bunch of things we need to plug into Cloud Foundry so that when things go right, great, but when they go wrong in terms of performance or any other issues, you can have visibility into what is going wrong and you have the ability to remediate it," says Chen.

Red Hat couldn't agree more, but adds that a platform cloud also needs to have hooks to integrated development environments, version control systems, and code build systems, and therefore OpenShift can have the Jenkins and Apache Maven code build management tools and Git software version-control system plugged into it. Red Hat will add support for more languages, frameworks, and services over time, as customers demand it, too. But Red Hat thinks it has another advantage.

"All of these PaaS services are built on Linux, but only one vendor has the Linux expertise and experience that allows us to really build out the capabilities," says Fernandez. "That is Red Hat, and I think there is more that we can innovate here."

But don't get the wrong idea and think that languages and frameworks will drive PaaS adoption. "I don't think it is strictly a feature thing," says Fernandez. "OpenShift is not just a platform with agility and flexibility, but has IT operations capabilities, too. We have had as many OpenShift conversations with the IT ops side of the house as we have had with developers."

In many cases, the ability to run a platform cloud internally or externally is important, and for those platform clouds that are going to be run in a virtualized environment, having an infrastructure cloud is also a prerequisite. (OpenShift can be deployed on physical servers because it uses Linux containers and SE Linux access control to isolate workloads, but you can also run it atop the KVM hypervisor or even Microsoft's Hyper-V or VMware's ESXi if you want.)

Data is mass, and the more data you have, it is harder to move it around

The location of the data used by applications is also a big determining factor – perhaps the most important factor.

"One of the things that we think about when we talk to enterprise customers – and one of the reasons why platforms will get adopted in the public setting versus the private cloud – is the notion of data gravity," explains Chen. "Data is mass, and the more data you have, it is harder to move it around. It is hard to move around petabytes, gigabytes, or even terabytes of information cost effectively or quickly from cloud to cloud.

“So as a result, I think that some of these public PaaS clouds will see some adoption for greenfield, net-new applications. But for any kind of application that needs existing data or legacy resources like that AS/400 box from the minicomputer era, enterprises are going to want to have a private cloud version of that technology so it sits closer to the database. That's being proven out by what VMware is doing with the New York Stock Exchange and others partners, who have large data sets that they want to start building clouds around."

The thing to remember is that this is just the beginning of the platform cloud wave, and the thing to do is get educated and experiment with various options. No one else is much further ahead than you.

"Right now, most of the PaaS adoption is at Web 2.0 startups and departmental edge cases," says Engine Yard’s Dillon. "It is shadow compute, mostly. The core IT guys are interested and they are sniffing around it, but it is anything but wholesale adoption.

“But if you are the CEO or board of directors, and you are hammering on your people to change the world, raise your top or bottom line, improve your customer loyalty or employee retention – you name it – you are not going to allow a three to four year waterfall development methodology and build a new data center. You are going to use information technology to innovate, and the infrastructure cloud, the agile languages, and the application platform clouds makes it possible to be all teeth and no tail. It's working so well that the core IT departments are going to have to develop this in some form or fashion." ®