Dave Ginsburg, CMO at Pluribus Networks, a US networking start-up that announced its EMEA launch today, does not like the label "software defined networking (SDN)".
"It's an overloaded term. There are still a lot of people with their own interpretation of what it means, and people are sceptical about whether there is any real revenue generation going on," he told Computing.
Ginsburg described as "noble" the goals of projects like Floodlight and Stanford University's Cleanslate in developing ways to virtualise the network, but he said those goals have become confused.
"What happened between the initial developments and the industry getting hold of it was a lot of rhetoric and start-ups, confusion, and protocols and different approaches to the point that you have to define what your definition of SDN is before you can start having a conversation about it.
"I talked to a potential customer for 20 minutes recently before the phrase SDN came up," he said.
Instead, Ginsburg prefers to focus on the problems that SDN-type technologies are designed to solve.
"It's network administration and management, the velocity of new services in a multi-tenant environment, being able to virtualise the network and create tenancy, being able to more effectively monitor and analyse the traffic within the network, and achieve full network visibility - all in a fully programmatic way," he explained.
Formed by ex-Sun engineers in 2010, Pluribus produces a "server switch" that runs a distributed network hypervisor called Netvisor. This allows the network to be managed and programmed very much as if it were a server, effectively consolidating the compute, storage, network and virtualisation elements, and allowing full visibility across all layers for applications, services and administrators: "converging NetOps and DevOps", as the company puts it. Pluribus is part of the Oracle Partner Network and has already formed working partnerships with Red Hat and Tibco.
Unlike with most SDN architectures there is no separate controller to interface between the underlayer (the physical network switches and storage) and the overlayer (the hypervisor and applications). Instead, a distributed network fabric is created between the switches with one switch in the fabric allocated controller duties. This, Ginsburg said, reduces latency associated with typical controllers, does away with flow table limitations and performance issues associated with the common command protocols such as Floodlight controller and OpenFlow. Data is transmitted through L2-3 without tunnelling through to L4-7 layers.
The switches contain SSD primary storage and PCI slots for flash, allowing for high-speed transmission between compute and storage, and for historic network forensic data to be stored as "time slices" where they can be analysed to determine the cause of any performance issues.
Netvisor allows for the creation of individual virtual network tenancies, allowing a service provider, for example, to create separate virtual networks for individual customers, including over wide area protocols. The programmability of the switch and the architecture makes it application-aware.
"Applications can make very fine-grained calls to the switches because of the control architecture between the compute and the switching," Ginsburg said.
One pairing that has already borne fruit is that with integration software firm Tibco, which has offered its Enterprise Messaging Service (EMS) as a standalone appliance based on the Pluribus server switch since 2012.
Ginsburg believes a key market for Pluribus will be large corporations developing their own private clouds, over 80 per cent of which he claimed - in the case of Fortune 500 companies - will look to OpenStack, the open-source cloud platform developed in partnership between Rackspace, Red Hat, HP, IBM among many other parties.
"In partnership with Red Hat we're consolidating the OpenStack controller environment into the top of the rack so it's no longer on separate server blades. Press a single button and the whole cloud rack will instantiate."
"It's an audacious goal but what VMware was to server virtualisation we want to be to network virtualisation," Ginsburg said.
Sometimes, the power of the mainframe is the most cost effective answer. Computing's Peter Gothard puts Computing's readers' questions on the future of the mainframe to IBM's Z13 expert Steven Dickens.
This Dummies white paper will help you better understand business process management (BPM)