Agility is arguably the most valuable ability that IT can give to the business. The ability to respond as quickly as possible to changing business needs and market dynamics is increasingly driven by technology and most specifically, fast access to high-quality data. The problem is that this data is being generated and stored across a variety of endpoints, edge and core data center environments, as well as cloud services. Not only are these storage environments heavily heterogeneous, but the networking environments are as well. More sophisticated orchestration via software-defined networking (SDN) is required to enable the right data to get where it needs to be, when it needs to be there, as cost-effectively as possible.
The relationship between storage and networking is becoming more complicated as data becomes distributed, as more storage and networking infrastructure controls and services are abstracted from hardware, and as multi-tenancy becomes more popular (for on-premises infrastructure as well as for off-premises cloud services). A growing number of factors contribute to the ability to get data from “Point A,” where it is stored; to “Point B,” at performance levels specified by service level agreements (SLAs), and IT teams are attempting to serve more users that are requesting access to more data. Naturally, available network bandwidth is an important and also a volatile factor in this equation.
SDN decouples the control and data planes, and as a result enables network resources that are distributed and heterogeneous to be controlled and orchestrated automatically, from a centralized management console. Data can be automatically routed based on network traffic patterns and available resources. However, IT professionals should be aware that many SDN architectures are still managed by an external controller. This means that either multiple controllers need to be deployed (which flies in the face of the simplicity and integration that SDN is attempting to achieve), or it results in the entire network being controlled by a singular controller (which may be an expensive and complex task from an architectural perspective, while introducing a single point of failure into the network).
To create a network architecture that can be tuned dynamically according to changing workload requirements more easily and cost-effectively, IT professionals should consider moving away from a centralized network controller. For its part, Pluribus Networks offers what it calls a “controller-less” architecture that distributes the network intelligence into the switches themselves, so that the network may be controlled from any switch. For further discussion, access our webinar, “The Software Matters in Open Networking,” on demand.