SDN and the New Economics of NetworkingBy Arthur Cole | Print
Re-Imagining Linux Platforms to Meet the Needs of Cloud Service Providers
SDN means commodity hardware can now be deployed and repurposed in any form that is desirable. The key is to ensure that open formats are allowed to thrive.
The last bastion of hardware-centric infrastructure, the network, is under assault. But as enterprise executives have seen with their server and storage facilities, the shift toward logical resources is a net gain in both cost and flexibility. As Infoblox’ Founder and CTO Stu Bailey notes in a discussion with IT Business Edge’s Arthur Cole, software-defined networking (SDN) means commodity hardware can now be deployed and repurposed in any form that is desirable – as a firewall, load balancer, what have you. The key is to ensure that open formats are allowed to thrive without giving ground to those who prefer the current hardware-centric way of doing things.
Cole: SDN has been in the news a lot lately, but there is still a lot of confusion surrounding it. In your view, what exactly does SDN bring to the table?
Bailey: The most important thing that SDN does is change the economics of networking. Instead of buying a dedicated firewall, load balancer or router, imagine buying a relatively inexpensive generally programmable network switch and deciding later whether it is a firewall, load balancer, router or more likely a combination of those things tailored to your needs based on the software which controls it. We buy servers and storage this way. We can deploy and repurpose them as needed and not worry about changing them often because we didn’t pay a lot in the first place. Software is the center of value. SDN is about bringing this value-added software and inexpensive hardware model to networking so your business can be more responsive and efficient.
The networks we create every day can become the source and foundation of unprecedented power and innovation with SDN. But the network is much bigger than switches, routers and boxes. It is defined by the relationships and importance of everything with an IP address that is connected and how it’s all communicating and interacting to provide a business service – from switches and routers to PCs and tablets to point-of-sale systems and RFID readers. SDN has the potential to create strong, secure, scalable, reliable networks that will support the radical network growth spike that is on the horizon with demands associated with cloud/virtualization, BYOD, Big Data, etc.
I am happy to see the growing interest in SDN, but the industry hype is beginning to create confusion about what SDN really is. True SDN puts software at the center of value, thus unleashing innovation, reducing complexity and radically changing the economics of networking. And the power of a software ecosystem will only be set free if we fix what is missing from the current networking industry: generally available, inexpensive, programmable hardware. This will mean that network managers can finally attain the economic and service capabilities of affordable hardware while benefitting from sophisticated, value-added software.
The architectures pushed by some of the hardware-defined networking incumbents like Cisco and Juniper might address the complexity challenge, but they do not change the long-term economics of the network and reduce cost like standards-based SDN has the potential to do.
Cole: Is it vital that the SDN ecosystem center around a single format like OpenFlow? Would we get the same benefits from a multi-protocol environment?
Bailey: It is critical for SDN to have a fully open way to program inexpensive network switches and easily turn them into firewalls, load balancers, routers and more. OpenFlow is the most mature standard that achieves that open commoditized programmability, and it has the advantage of being driven by network buyers instead of any one vendor.
Agreeing on an industry standard is the only way we will be able to realize the full potential of SDN and put software at the center of value in networking. Software has practically unlimited promise when coupled with generally available, inexpensive, programmable hardware. Take, for example, the remarkable innovations that were born from software on standardized x86-based hardware like servers and PCs – the Web, the cloud, etc.
There is no single preferred standard yet, but OpenFlow is currently the best way to get there. If SDN evolves and we see multiple standards emerge, we will not get the same benefits of a single protocol. There will be a rich software ecosystem, but there will absolutely be complexity challenges for consumers. Think of it like managing an environment with SPARC and x86 servers – it’s possible, but complex. Automation will be key in this scenario.
Cole: There are tales of applications being embedded with their own SDN-based network configuration modules, allowing them to craft complete operating environments for themselves. Do you think the enterprise is ready to give up that kind of control to software?
Bailey: I see this transition as inevitable. As SDN develops and matures, the market will demand an SDN-embedded network configuration module. Eventually, I envision a network that could scale for SMBs and multi-national corporations and service providers alike. With security as a primary concern for network managers, businesses need to transition toward a service that can offer changes, additions to the capacity, and enhanced functionality, all at the same time and in the same place. SDN opens up an opportunity to operate the network itself, instead of its individual elements.