How to define “best practice” in a quickly evolving cloud ecosystem
Radical challenges encourage radical solutions, and the CloudEthernet Forum (CEF) has decided to develop standards in a way that is very different from the traditional standards body approach.
Will the CEF be a standards body, and will it announce a certification program were among the first questions raised when we launched the CEF last year, but it was too early to give a definite answer at that stage. We were faced with a set of critical issues and it was up to us to first define those, and then agree the best strategy to resolve them.
The CEF began life in listening mode. Tata Communications was among the first to note nascent difficulties emerging when their bigger customers began to seriously scale up their datacentres and take them global. We had our ears to the ground for early signs of impending thunder at a time when a wider public was being confidently assured that the cloud was one hundred per cent silver lining. CEF founder members agreed that in time the cloud would be the promised solution, but serious co-operation would first be needed to resolve these issues before the cloud scene fragmented into proprietary silos.
A year later, and our mission is already a lot clearer. When the MEF, for example, began developing standards for Carrier Ethernet, their task was to persuade the industry that Ethernet was a WAN solution, and creating standards was an essential part of that mission. Our situation is very different, because cloud services are already forging ahead at full steam requiring sound rails to be laid ahead to make sure the market does not come unstuck. To achieve this we must integrate three relatively new concepts: NFV, SDN and Carrier Ethernet. That’s uncharted territory, it is highly complex, and it is needed under the pressure of rapid cloud deployment.
Cloud computing is as radical and disruptive as the arrival of the personal computer. What happened then was an influx of new personal computers with a whole range of incompatible operating systems. The market eventually shook down with the dominance of Microsoft and the IBM PC, but there were still incompatibilities with Macintosh and Unix, and it was generally recognised that the dominant OS was not necessarily the ideal one for business.
This is what we now risk with cloud computing: there are very good cloud services emerging, but they are not designed to global standards. Again businesses will face the problem of trying to integrate its global operations while depending on incompatible proprietary services, however good those services might be. Industry standards will be needed to build the sort of vendor-neutral cloud services that are required to support a truly vibrant global business culture.
That is the objective of the OpenCloud project announced in May. The CEF has named its emerging set of standards “CloudE 1.0”, where the “1.0” is a deliberate tag to suggest an evolving standards development process that is radically different from the norm.
Most standards bodies work on something like a quarterly cycle. They meet and debate towards a consensus on what the required standard should be, then draft versions are shared around and refined until the next version is agreed and approved at the following quarterly meeting. This approach would not work for something as complex and fast moving as the cloud, so CloudE 1.0 will be developed as an on-going iterative process based on a feedback loop involving a Cloud Reference Architecture, a Reference Test Bed (the OpenCloud platform) and a growing set of Use Cases based on real business needs.
The “Cloud Reference Architecture”
“What exactly is the cloud?” sounds like a naïve question, but it is still being asked. We chose to define our cloud not in abstract terms but by creating a reference architecture containing all the essential components – the interfaces and functions as well as the players – and then building it in real life.
One of the first benefits for CEF members is the opportunity to shape this “Cloud Reference Architecture” by contributing elements to it. This casts the membership net very wide: not only NEMs, CSPs, data centre operators and carriers, but also any enterprise anticipating becoming a cloud consumer can become a player, and can now get its hand in to help shape tomorrow’s cloud environment to its needs. Membership is already expanding in response.
The Reference Architecture is already well established, but it will continue to be an organic, growing structure – both in response to new developments and as part of an iterative process.
The “Reference Test Bed”
Iometrix president, Bob Mandeville, is responsible for the creation of a Reference Test Bed that will be very different from a “proof of concept” exercise, as he explained: “Proof of concept is a short term process, whereas we are creating a permanent test platform that mirrors the CEF Reference Architecture and will evolve over time. We are not testing a single concept, but how the reference model responds to realistic use cases reflecting CEF members’ actual business needs”.
A very simple test scenario could be a virtual machine representing the initial state, then the virtual machine is relocated, or else multiplied into one hundred thousand clones, then finally returning to a single VM for the final state. Can this happen while preserving all the VM’s attributes and functions, with the required logical resources (such as connectivity) provisioned automatically? The results of each test feeds back data for refining the reference model until it is considered to be a suitable basis for defining an industry standard.
While the underlying architecture is already taking shape the test bed is very much work in progress, with the CEF working with members to supply equipment to represent each component of the architecture. This provides another interesting opportunity for members – to get a head start by seeing how well their products will perform in a leading edge cloud environment.
That is how the CEF will evolve its role as a standards body: reference architecture to test bed; impose use cases; note results and feedback to the reference architecture until is seen as a sufficiently stable basis for the next release of CloudE standards.
Use cases are already defined for application performance management, cloud security and traffic load balancing, all across multiple providers – with a growing membership keen to suggest their own challenges needing to be resolved. This is very much in keeping with the forum’s aim to stay highly relevant to real business needs and not become an ivory tower exercise.
In paving the way to the definition of standards for an OpenCloud environment, the CEF is shaping a radical solution to a radical problem.
Cloud services rely on the end to end interoperability of so many players – enterprises, network and data centre equipment vendors, data centre operators, orchestration layers, management and reporting platforms, security devices, network service providers – the list goes on. While the MEF has a successful model of defining service types and attributes that everyone can agree to and align with, the CEF aims to bring a similar impetus to the cloud industry by adopting an iterative standards development process to keep up with the rapid evolution of the cloud ecosystem.
Unless we can define industry best practices and global standards to establish an open cloud environment, cloud services will become more and more fragmented and difficult to integrate. A rapid, iterative and above all business-focused process is needed to ensure that interoperability between components is made as standardised as possible. This then enables true innovation and differentiation at a service layer, and with that come new revenue opportunities.
By James Walker President of the CloudEthernet Forum