Avoiding standards lock-ins
Today's SDN "definition war" is a disturbing trend impeding cloud service providers from implementing practical and useful SDN products that provide granular control to customers. SDN's scope and definition have diminished to the extent that one cannot differentiate it from the term OpenFlow. If these definitions are convoluted and misappropriated by the industry, the useful and dynamic purpose of SDN will become overlooked.
SDN might get the attention of some system administrators, who want to perform traffic engineering and reshaping, but the momentum this new technology brings to network device makers and users alike will begin to subside. After many years of suffering from shortsighted networking protocols and the need for expensive optimizers, accelerators, and adapters to handle their inefficiencies, customers want straightforward products that provide interoperability, granularity, dynamism, and greater visibility. Standards are meaningful only if they provide true value and are extensible.
As customers try to avoid the "vendor lock-in pit" in hope of realizing a cost-effective IT solution, they should not be diverted into the "standards lock-in pit," either. If the propagators of SDN terminology wish to confine their definition to a smaller and simpler southbound protocol like OpenFlow, then it will become just another piece of networking jargon that fades away.
What's needed is SDI, an intelligent and automated control plane that covers the breadth and depth of the IT infrastructure — and decouples and unifies the control plane to provide holistic value across the data center. For example, an "SDN" controller should be able to dynamically control everything from HVAC systems to Layer 0 devices such as ROADMs (reconfigurable optical add-drop multiplexers) and everything in between.
Asking the right questions
SDI should enable cloud service providers to ask the right questions to define the control plane parameters for their customers. A true SDI solution should enable the cloud service provider's provisioning interface answer the following:
- What is your daily/weekly/monthly budget per service?
- How many users are you expecting for each service at specific times?
- How many maximum user sessions you would like to provision?
- When the maximum budget threshold is reached, which services would you like to remove from prioritization, so you can redirect those resources toward specific higher-priority services?
- Do you want an automatic service scaling or approval-based auto scaling of services?
- What are the security and administrative restrictions on service consolidation and migration? (Some service customers want to run sensitive services on a private cloud, while the rest run on a multitenanted cloud.)
Implementing true SDI
The ultimate goal of a correctly implemented SDI solution is to provide an autoscaling services infrastructure that's not grossly overprovisioned, thereby optimizing the cloud customer's operating expenses — while ensuring full operation during periods of expected or unexpected heavy use.
Sign up for Computerworld eNewsletters.