Cloud’s rise as a pervasive platform

Buzz Lightyear might disagree, but cloud is a platform, not a destination.

Notice that I did not say "the cloud" but rather "cloud." That differentiation is important because I'm talking about a thing -- not a place.

New and widespread computing platforms appear only every so often, and a set of cloud technologies in aggregate represents the next pervasive platform. "The cloud," on the other hand, is clearly a place that is an implementation of a cloud platform, plus a whole lot of process and procedure to make it run.

While the place is certainly interesting and yields some great marketing, it's the platform (i.e., the thing) that makes it all happen. Without a true platform that enables it, "the cloud" is really just an outsourced data center. Need more convincing? For perspective, let's take a quick look back at some previous platform shifts, such as the dark days when we had to tweak our NOSes to create a (somewhat) usable PC server platform.

Tweaking our NOSes

In the late 1980s, the first network operating systems (NOS) started to appear. Their primary use case was file- and print-serving. The way you deployed one of these was to contact a network reseller and pay a ridiculous amount of money for someone to come install network adapters and device drivers in desktop PCs and to also install the NOS on one larger desktop PC. For the server and a few desktops, software alone might have run into the $7000 range, and then the cost of labor was more. Hardware was very specific to the NOS and the desktop PC drivers. To create a file server, what the NOS vendors and their high-margin resellers did was bend the desktop PC platform (something built to run character-mode word processors and spreadsheets and do dialup communications). As a result, performance and reliability were hard to characterize, because the platform was not built for the use case.

Birth of the PC server platform

In the early 1990s, the hardware guys decided to create an adaptation of the deskside PCs that was properly configured to suit the server use case "out of the box." That is, networking was standard on the machines, as were properly sized and performing memory, cache and disk subsystems. The new platform was tested with new NOSes (like Netware, OS/2, Windows NT, and a bit later Linux and ESX) that were designed with the target platform in mind, as opposed to being retrofitted. Total cost of ownership came down dramatically, and the NOS became commodity-priced. The PC server platform was born and was defined by both hardware and software functions.

Of course, I'm oversimplifying all of this, but I recall history in order to set context for where we are today.

Cloud formation

The venerable operating systems of the 1990s are at the core of what runs applications in the cloud today. So EC2, Rackspace, Joyent, and the various VMware and Windows Server hosters use those operating systems -- coupled with hypervisors, modern chipsets and the server design work that was done -- to create very stable and scalable server nodes. On top of and underneath those systems are management layers that have been put together to express the server utilities we recognize as familiar infrastructure as a service (IaaS) today.

While important and necessary, the natural evolution of those operating systems is not "cloud, the software platform." When you take one of the server OSes from the 1990s and the 2000s, along with its captive set of server applications and tools, and try to say it is a cloud platform, you are tweaking it in the same way the early NOS guys bent the PC platform to make it act like a server. There are new systems like Windows Azure, which has taken a fresh approach while starting from a known core.

What now?

I'm not saying we should throw away the OSes and hypervisors that have taken us this far. Those structures do an amazing and all too easily forgotten job of making the server nodes reliable, scalable and secure. What I am saying is that the cloud platform is something much more. This is something designed for a different kind of scalability, management (or lack thereof), application enablement and security that can be layered onto the core that works so well today.

In this inflection point, we can think about products and services that were not possible before. Recently much of that action is in the layers above the OS and hypervisor (an example is VMware's recent Cloud Foundry announcement). The folks at Amazon (AWS) and Microsoft (Azure Services) are working at it too, in addition to a bevy of startups pushing the envelope and challenging traditional thinking. Red Hat recently entered the space to add more to the mix with its OpenShift Platform as a Service.

This blog will be about the platform and the requirements it aims to satisfy, with news and opinions and stuff in between, and I'll draw heavily from patterns that I have observed over the past 26 years. Stay tuned!

Frank Artale is a partner at Ignition Partners, a venture capital firm, where he specializes in cloud, core infrastructure, networking, and security investments.