In his blog post (from Nov 2011), Alvaro Retana from HP is making the case that SDN / OpenFlow is just another Network Management solution. In his conclusion, he writes:
SDN is a step in the right direction—toward making the network more dynamic and self-adaptable, but it currently represents the evolution of network management, not the evolution of networking.
I would agree that it’s true that SDN1 is better network management than we have today (most of which is a wasteland of disappointment). I believe that today’s Network Management is achieved using sub-standard protocols that have poor APIs. Network Management is based on the use of certain protocols to provide input data to the management platform and for networking. Those protocols are SNMP, Telnet/SSH and maybe some RMON or SNMP traps. These protocols are very limited in their capability. Although SNMP & RMON were designed in a different time when CPU and memory were constrained and device soft
Vendors have consistently demonstrated an unwillingness to adopt new protocols such as NETCONF/YANG that might expose richer data and information. The value of OpenFlow as part of Software Defined Networking is that it can combine with existing protocols to finally provide meaningful impact to the configuration and operation to the data plane.
Current network management platform strategies, as typified by HP OpenView platform, have consistently failed to deliver any value to networking. Indeed, any mention of HP OpenView to Network Engineers will result in laughter, outright scorn and horror filled war stories. And yet, OpenView continues to be sold to management & executives. We do need better Network Management tools, but those existing products such as Cisco Prime/CiscoWorks and OpenView are clearly proven to NOT be an answer.
I wrote previously about Network Hierarchy of Needs and how our networks evolve from simple to complex, at the same time moving from dumb to growing awareness and intelligence. To my mind, Network Management of DATA Networking has not been delivered by any company. Our existing tools are not good enough, or surely one of them would have developed something worthwhile by now.
Therefore, I reject his proposition that OpenFlow/SDN is only about Network Management, and believe that it offers much more because the underlying tools give us much more capability and functionality. Rich tools and APIs deliver rich applications and capabilities and that’s what Networking needs to meet the virtualization challenge over the next five years.
I think that Alvaro might be trying to visualise the future using the frame of yesterday’s legacy tools, such as HP OpenView. And that’s a sure path to avoid innovation. In my view, HP needs to either throw out OpenView because it’s static, hard to use, poor tools, and lack of flexibility or, at the very least, significantly change that Business Units approach to adapt to the next generation of dynamic networking. The OpenView team have never managed to deliver anything significant to Data Networking in the last fifteen years.
I’m confident that the future of SDN isn’t going to look like HP OpenView product strategy (or BMC or Tivoli for that matter) because that’s yesterday’s model for technology automation. Tomorrow’s vision of network operations with orchestration and automation will be quite different.
As Alvaro writes in his second last paragraph
SDN is a new networking architecture. It is not a complete clean slate, but a new start. SDN represents a time to innovate and optimize, to take full advantage of all the industry has learned. But SDN is also at a point where the basic Infrastructure needs are being addressed: connectivity, network management, routing. ……
Sadly, that probably won’t stop the big vendors from trying to do it anyway and attempt to make their legacy platforms use SDN or even become “Cloud Ready”.
- SDN = Software Defined Networking ↩