Much debate has occurred recently over whether Amazon’s API should be considered a de facto standard by cloud computing service providers. Those who reject this notion (ahem, Lew Moorman) say you can’t clone Amazon’s cloud by copying its API alone.
“Having API compatibility for the basics like loading and getting an object is very easy but these technologies do complex things and will always be different because there are different technologies underneath,” Moorman, president of Rackspace, said in an interview for our July 10 Leaders of the Open Cloud article.
Cloud platforms must start from scratch to innovate and truly meet the needs of customers and vendors.
So I was intrigued when HP’s Richard Kaufmann told me in an interview a few weeks back that there’s one unique case study, in particular, that demonstrates how the OpenStack community is reinventing cloud computing – not just remodeling existing platforms.
The Quantum Project is working to design next-generation virtual networks with no model for comparison, said Kaufmann, chief technologist for cloud services at HP, which contributes to the Quantum Project.
I heard much the same statement last week from Lew Tucker, Cisco VP and CTO of cloud computing, in his OSCON session about the Quantum Project. This is new ground. There’s no such thing as programmable infrastructure, or what project leaders call a network-as-a-service. The Quantum Project has set out to correct this, starting by problem-solving the issues with current technology.
“Fixed networking today is really a problem,” Tucker said to a roomful of sysadmins and developers.
“When you look at the general model of cloud computing, networking is tied up in the compute service. That makes it really hard to build that service,” Tucker said. “Provisioning of various vendors is hard and maintaining will be even harder.”
The Quantum Project aims to break out networking from the rest of the OpenStack services to build a new layer of abstraction for two-way communication between an application and the infrastructure.
“The network service completes the trinity of the cloud – with compute and storage services,” Tucker said.
Quantum, a core part of the OpenStack project starting with the next Folosm release, aims to allow the software above the stack to control that infrastructure through a well-defined interface – an API that developers can access. Project participants are also developing a plugin component to the architecture to allow customers and vendors to experiment with new extensions but keep the API stable.
“We hope to nail down the API very shortly,” Tucker said.
He predicts this mechanism of a simple API with plugins will lead to the development of many vendor-specific plugins for things such as quality of service and port profiles. A slew of new networking services will also arise as a result.
“We’ll have a new way of inserting computing into the network,” he said. “A lot of things around data flow will start to drive it. And we’ll end up with a new notion of what we mean by a virtual datacenter.”
The Quantum Project has big aspirations and it hopes to meet them by working through a community process in an open format so that all can contribute and benefit.
“OpenStack is really setting out to build an alternative end to end,” Moorman said. “When they say we’re going to do networking, they’re not just setting out to copy Amazon, but asking how should networking be done in the modern world?”