Subscribe / Unsubscribe Enewsletters | Login | Register

Pencil Banner

BLOG: Software-defined infrastructure: Cloud services do SDN one better

Karthikeyan Subramaniam | Oct. 3, 2013
Adara Networks has been selling SDN equipment to service providers for years. Adara's chief software architect explains how SDN's advantages extend across the data center

Adara focuses on software-defined networking and infrastructure orchestration. Karthikeyan believes we aren't looking quite far enough in terms of what can and should be automated and orchestrated across the data center.

Rather than limiting SDN to the network, we should be looking to deploy software-defined infrastructure, or SDI, that would separate the control plane for every aspect of the infrastructure, and lead to new ways to manage infrastructure resources and costs. — Paul Venezia

Running cloud services within budget using SDI
True plug-and-play IT infrastructure is fast approaching. With the advent of cloud computing, many small and medium-sized businesses may soon be able to rid themselves of IT infrastructure maintenance once and for all.

SDN (software-defined networking) is a vital component of effective cloud computing for both the provider and the customer. In fact, it should involve more than just the network — and extend to all aspects of the infrastructure, from the data center environment to server processing resources. This expanded definition might be better termed SDI (software-defined infrastructure).

It's all about visibility and granularity
In an IT infrastructure based on physical servers, one of the main issues is a lack of visibility. When accounting for IT and data center costs, a large percentage of capital expenditures are actually allotted to bandwidth, storage, and other operational resources. To be properly assigned, these operational resources need a correct method for planning and eventual deployment. Otherwise, these resources may become misallocated — wasting time and money, potentially on a large scale.

Virtualization by itself is not enough. A hypervisor does not necessarily lend visibility into what services are running inside a virtual machine. Often, overhead within an operating system may consume significant resources. A primitive automation system, for example, can mistakenly spin up another virtual machine to offset the load on enterprise applications that are presumed to be overloaded.

SDN offers the potential for greater visibility and granularity by separating the control plane from the data plane. Academic and research communities have created software components called controllers to serve as the control plane, which provide API or CLI-based interfaces to push rules to switches and manage data flow. These tools are generally too complicated for administrators to learn, program, or manage. Not many administrators are interested in becoming programmers.

A production-ready control plane needs to be simple, yet intelligent enough to parse business SLAs at a granular level and provide the automation needed to enforce them. It should allow system administrators to allocate the appropriate resources with service-level granularity. All protocols, mechanisms, and tools should provide this granular level of control, in contrast to control planes that are tied to physical hardware. SDI unifies the data that's scattered from physical to application layers and illuminates granular, service-level resource usage and control throughout the infrastructure.


1  2  3  4  Next Page 

Sign up for Computerworld eNewsletters.