The IT community has a pretty good track record when it comes to changing data environments for the better. This is part a function of technology and part simple economics: good ideas get rewarded with implementation and revenue streams while bad ones fade away.
Lately, though, it seems that the advancements are coming so fast that the enterprise barely has time to give one a full examination before the next one is at the starting gate. A key example is software-defined networking (SDN), which is already shaping up to be the next big thing even while most organizations are still trying to get their heads around the cloud.
And while it may seem that the concept of separating network hardware from network operations with an abstract software layer is pretty straightforward − after all, it did wonders for server infrastructure − the fact is that there are a number of different flavors of SDN that offer both unique capabilities and varying changes to existing network hardware.
At the moment, Cisco and HP seem to be squaring off over the initial SDN market. Both were at VMworld last month with SDN solutions aimed at leveraging VMware’s considerable virtual footprint in enterprise circles. Of the two, it seems Cisco has fostered the more formal relationship with VMware, signing a deal that integrates the Nexus 1000V switch into the vSphere 5.1 management suite even as it incorporates vSphere and vCloud into Cisco’s overall Unified Computing System (UCS) portfolio.
But that isn’t to say HP is left out in the cold. The company is touting a range of new software platforms, including the Ethernet Virtual Interconnect (EVI) network overlay solution as a means to not only virtualize networks, but streamline them as well. HP also lays claim to more faithful support of the OpenFlow protocol, which should prove more interoperable than the “optimized” version that Cisco is said to be developing.
Still, HP and Cisco are not the only ones with hats in the SDN ring. Dell is hard at work on OpenFlow systems aimed at coordinating SDN activity all the way to the hypervisor. Most implementations to date involve the controller and the data plane − switches and routers mostly. Under Dell’s approach, kicked-started largely through the acquisition of Force10 Networks earlier this year, enterprises would be able to streamline the SDN management stack, offer a way to integrate all leading hypervisors into the fold and provide better coordination between dynamic network topologies and rapidly shifting data loads. It also would provide Dell with a continued stake in the game as SDN assumes greater responsibility for network performance and configuration tasks and leads to the commoditization of network hardware.
This same tack is also playing out at network-centric firms that don’t have lucrative integration deals with VMware. Brocade, for one, has invested a lot of resources into its network fabric solutions and doesn’t want to see it outclassed by SDN. To that end, it has teamed up with Microsoft to get in on the ground floor of Windows Server 2012. Brocade will support the new OS on both its Fibre Channel and Ethernet SAN platforms and has updated them with tools like the Dynamic Circuit Network bandwidth allocation solution and the Network Virtualization Using Generic Routing Encapsulation (NVGRE) overlay networking protocol. Again, the name of the game here is streamlined network architecture that can be both easier to manage and provide a more dynamic environment for virtual and cloud operations.
The very notion that SDN will allow enterprises to replace their high-tech routing and switching infrastructure with commodity hardware is bound to make more than a few CIOs crack a smile. But placing all of that network intelligence on the software layer is still largely an untried and untested practice.
There’s no reason to think that it cannot be done − in fact, Cisco, HP, VMware and many others are more than willing to bet big dollars that it can − but there is nothing like real world experience to take a true measure of the drawing board calculations.
Only then will we know whether SDN has true staying power.