Do you like this site? Remember to share it to all your friends on Facebook and Twitter!

Sunday, December 8, 2019

Three possible scenarios of Software Defined Networking (SDN)

Software-defined networking (SDN) is an approach to create a centrally controlled programmable packet network. Any protocols with the same approach could be considered as SDN as well.

For open protocols, we have one popular standard protocol “OpenFlow” talking among the central controllers to all managed networking devices. Open Network Foundation (ONF) defines OpenFlow protocol.

In fact, vendors also have developed proprietary protocols to implement this same approach. For example, Cisco’s ACI is a proprietary SDN solution.

Here I summarize 3 most probable scenarios when we deploy SDN.


Scenario 1: Open protocol, open multiple vendors


Since OpenFlow protocol from ONF is open, any vendors can develop inter-operable software and hardware products. For enterprise customers, the first natural approach is to buy from multiple networking venders.

For example, controllers from vendor A, some switches from vendor B, some routers from vendor C, and so on and so forth.

The most obvious benefit of this scenario is lower buying cost. Enterprises can buy any compatible networking products from any vendors in the market with the lowest price. White-brand, or no-brand vendors have opportunities to compete on price against existing networking vendors.

However, only the buying cost is lower. We also must consider other costs to build and maintain a working network. Integration of software and hardware itself is a heavy project.

When we already have a capable team of hardware and software integration, we can work comfortably with this approach. If we simply don’t have such a “Tiger Team”, or we are just about to create a team from scratch, this scenario could be difficult and costly. It could cancel out all benefits of lower buying cost.

Scenario 2: Open protocol, one major vendor


Some vendors are capable to provide all components for OpenFlow. For example, Cisco. In this scenario, basically we buy controllers and network devices from single major vendor. For less important areas, we buy some from other venders in the market.

In this approach, we might have higher buying costs. Because we now have a major vender, we can gain better support from that major vendor. We can also achieve less integration cost because our team have fewer combinations of products to experiment and integrate with. We don’t need a huge team like previous scenario.

I am more familiar with Cisco. Let me summarize what Cisco can provide for OpenFlow.

“Cisco Open SDN Controller” is OpenFlow protocol controller. The software is a commercial distribution of OpenDaylight by OpenDaylight open source project. This software is packaged as a virtual machine format.

In addition, Cisco’s Nexus 3000 and 9000 family switches can run “Cisco OpenFlow Agent” inside to become OpenFlow switches so they can be controlled by standard OpenFlow controllers.

We can deploy OpenFlow by simply selecting all components from Cisco. Because OpenFlow protocol is open, we also have the flexibility to add non-Cisco but OpenFlow compatible devices.

Scenario 3: Close protocol, one vendor


Some vendor can provide all features and benefits of “centrally controlled programmable packet network”, with proprietary protocol. For example, again, Cisco.

Cisco’s Application Centric Infrastructure (ACI) is Cisco’s proprietary SDN solution. With Cisco’s ACI, we can achieve even more than OpenFlow such as:

  • Device management
  • Better integration with non-networking devices such as Layer 7 switches and stateful firewalls
  • Better programmer-friendly abstraction instead of VLANs and subnets.

In this scenario, we have the highest buying cost and we are locked into single vendor. However, we have the lowest integration cost and we now have full support from that single vendor. We only need an even smaller support team and concentrate all resources on using the network instead of experimenting interoperability among vendors.




One more thing…


Winter flowers near Taoyuan High Speed Rail Station.


SDN is a promising approach for next generation networking. Programmable network indeed is the foundation for network automation.

On the other hand, I don’t think it fits well for all types and sizes of customers. Let me talk more about who needs SDN in the coming posts.

I am Li-Ji Hong. This is my blog “Show IP Protocols”. See you next time!
Do you like this post? You really should consider Subscribing by Email!


Related Posts with Thumbnails

No comments:

Post a Comment

Tip: you can also anonymously comment here.

Popular Posts