There is no shortage of groups working on cloud standards, but there are questions about whether enough thought has gone into the fundamental direction and impact of standardisation, says Lori MacVittie.
Mention virtualisation and most people immediately think of operating systems. That's understandable because operating-system virtualisation is at the heart of so much debate and so many major initiatives and shifts in datacentre models.
But long before operating-system virtualisation, we had many other forms of the technology, such as network virtualisation in VLANs and network address translation, and server virtualisation, such as proxies and load-balancing.
All these virtual technologies are still heavily relied on for reliability, high availability and fault tolerance in networks that support today's most complex, challenging datacentre architectures. Suppliers are adding more virtualised systems all the time to provide cloud-computing customers with choice and control over cloud-based deployments.
Costs of complexity
Virtualisation can introduce complexity, and the costs of managing that complexity can quickly offset the savings associated with consolidation. Fortunately, various groupings are attempting to combat this complexity. Unfortunately, they are not collaborating.
It would take up too much space to itemise every cloud and virtualisation standardisation effort. Between DTMF, OCCI, OGF and SNIA, there are enough cloud-standard initiatives to make the head swim. A good overview of these efforts can be found at the Cloud Standards wiki. But be warned: you are unlikely to find a promising picture of progress.
The goal of standardisation is two-fold. On the surface it is about portability and interoperability and, more immediately, ease of use. A unified cloud application programming interface (API) would allow consistent management across various implementations.
Under the covers, portability and interoperability also matter. But more immediately, standardisation is about infrastructure collaboration. It is infrastructure 2.0-focused efforts that will be most beneficial in the short term because the presence or absence of standards describing how infrastructure integrates and collaborates directly affects the efficiency of an organisation's datacentre. That situation holds true for both the provider and the enterprise.
As more of infrastructure becomes virtualised, the need for standards becomes more urgent. The architectures that include virtual network appliances necessarily become more complex. Standards would help deal with this issue by providing a common management and integration interface.
Part of the problem with cloud standards is the multiple layers at which standardisation is necessary. It is not enough to develop a cloud API for managing a specific environment from the top down, providing a necessary operational view. Perhaps more important is the ability of the layers of virtualisation within the cloud to communicate and collaborate in a standard way.
There is very little ongoing effort in this area. Management at the virtualisation layer via APIs and standards are the focus, with the ability to manage and integrate virtualised infrastructure software summarily dismissed. That emphasis may be a good thing.
That is because the downside to standards is that they are prescriptive. Rarely are they designed for the constant evolution of today's environments. Standards now might inhibit the maturing process of virtualisation and cloud computing.
Lowest common factor
Standards can also, sometimes unintentionally, reduce architectural choices to the lowest common factor. Standards must take into consideration a variety of systems in the problem area, so what results is something standard — that is, common across all implementations.
It is likely that cloud computing will indeed drive infrastructure towards what could be called the least common architecture. But primarily this would occur through the standardisation of infrastructure services at the API or control-plane layers.
If a standardised set of APIs are developed to support all vendors in a given architectural tier, then certainly advanced or innovative features or services will probably be excluded.
Without standards, each provider and vendor essentially has to go their own way, and it is from these efforts that standards will eventually spring.
Lori MacVittie is responsible for application services education and evangelism at application delivery firm F5 Networks. Her role includes producing technical materials and participating in community-based forums and industry standards organisations. MacVittie has extensive programming experience as an application architect, as well as in network and systems development and administration.