SDN Openflow commercial applications – Part 2

This blog is a continuation from my previous blog on SDN Openflow commercial applications. In this blog I will cover SDN applications in Network performance optimization and monitoring and Data center fabric.

Network performance optimization and monitoring

Networks have a lot of information and by looking at Network state and statistics collectively rather than at individual device level provides a lot of useful information. This information can be used to optimize the network based on application needs. The next set of applications are targeted towards monitoring Network state and statistics as well as improving the performance of the network.

Kemp SDN Adaptive load balancer

Kemp has integrated its Load master ADC with HP VAN SDN controller to provide adaptive load balancing based on the traffic load on the switches.  Normally ADC makes load balancing decisions based on L4-L7 characteristics. Kemp’s Load master ADC interacts with HP VAN SDN controller which in turn gets the traffic load on switches using Openflow. Based on the L2/L3 network statistics and the L4-L7 characteristics, Kemp’s Load master makes intelligent decisions on traffic steering and load balancing.

Realstatus Hyperglance

Realstatus’s hyperglance is a cloud visibility and management platform. Realstatus is integrated with Opendaylight and HP’S VAN SDN controller and it shows the network topology and flow related information in realtime. It is possible to monitor flows, identify hotspots, find the shortest path between 2 hosts as well as modify the flow path in real-time. Also, Realstatus provides 3D visualization of the network. Following picture shows a sample topology in Realstatus along with traffic flow between 2 hosts.


Ecode Evolve

Ecode Evolve application is targeted towards Network design, monitoring, troubleshooting, simulation and performance optimization. The Ecode application runs on top of HP VAN SDN controller. Following is a block diagram of Ecode Evolve application platform.


  • Evolve canvas is used for network design.
  • Evolve control takes a snapshot of the network on top of which network simulation can be done. This can be used to predict how the network would react under different stress scenarios.
  • Evolve netlyzer is used for realtime analysis and troubleshooting.
  • Evolve flow rector is used for traffic engineering.
HP Network optimizer

HP Network optimizer SDN application integrates with Microsoft Lync to improve application performance based on the application type and need. Lync is a collaboration application from Microsoft. HP Network optimizer works on top of HP VAN SDN controller which in turn talks to the Openflow switches. When network optimization is turned on, HP Network optimizer programs the Openflow switches to prioritize Lync application so that congestion and latency can be minimal for this application. This allows users of Lync application to see changes in realtime. Following block diagram shows the different layers of the solution.


Data center fabric

Typical Data center fabrics have a 3 tier network(Edge, Aggregation, Core) and they are mostly focused on North<->South kind of applications. Newer data centers have majority of their traffic in East<->West directions and Leaf, Spine 2 tier architectures are most suited here.  Typically, Cloud orchestrator like Openstack provisions the data center infrastructure that includes compute, storage and networking. Networking piece is handled by Neutron plugin which in turn talks to SDN controller to provision the network fabric. Following are examples of some SDN applications that automates the provisioning and monitoring of Data center fabrics.

Big Switch Big cloud fabric

Following picture shows the block diagram of P+V Big cloud fabric.


  • The components of this solution includes baremetal switches that support Openflow which can be used in Leaf and Spine, Switch light OS that runs on Leaf and Spine switches as well as in Vswitches and Big cloud controller which is the SDN controller.
  • Big cloud controller interacts with cloud orchestration software like Openstack and maps the logical networking topology into the physical topology consisting of Leaf and Spine switches.
  • Logical routers are implemented in the Leaf switch and in the Vswitch and load balancing is done between Leaf and Spine.
  • Overlay approach is not used here. Its still not clear to me how issues like VM mobility and overlapping IP address issues are addressed. I think VM mobility issue gets resolved because of having logical router at the Leaf.
NEC Programmableflow Openflow based Network fabric

NEC’s network fabric solution consists of switches from NEC that supports Openflow, NEC’s PFC SDN controller and virtual switch for Hyper-v hypervisor. Following is a block diagram of their solution with hyper-v integration.


  • NEC’s network fabric solution uses the programmable flow based model where the flow classification happens only on the edge of the network and packets gets switched in the middle of the network. This is similar to how MPLS switching work.
  • To provide isolation between multiple tenants, NEC has proposed the VTN(Virtual tenant network) model where each tenant gets their own virtual switch and virtual router. The virtual switches and routers gets mapped into the physical network by the PFC controller.
  • Most of the heavy lifting happens in the network edge where both the switching and routing functionality is done and minimum packet processing happens in the core of the network.
  • NEC’s VTN multi-tenant solution is also integrated in Opendaylight controller.
HP VCN(Virtual cloud networking)

HP VCN SDN application working together with HP Helion Openstack and VAN SDN controller provisions the Data center fabric. Following shows the block diagram of their solution.


  • This is an Overlay solution using VXLAN technology. The switches are programmed using Openflow.
  • For baremetal applications, HP’s physical switches initiates and terminates the tunnel.
  • VCN SDN application has enhanced Openstack Neutron to provide advanced networking capabilities. Following table from HP shows the additional capabilities of VCN SDN compared to Neutron.


Final thoughts

  • For a long time, Networking was a closed black box where hardware, control plane, management planes were all vertically integrated. Hyperscale data center vendors like Amazon, Google, Facebook have shown a lot of innovation by separating the different components. This model also allows newer players to enter this space with limited resources which allows for more innovation.
  • Openflow along with SDN controller exposes the capability of network switch through REST apis. This allows new network applications to be developed at a fast pace. Applications can be developed by Cloud service providers, Enterprises, End users based on their need without waiting for the network switch vendor to develop applications.
  • The common theme with all applications highlighted above are making the network programmable and getting the best out of the network.
  • Recently, there is a push from traditional switch vendors to expose their device APIs and to encourage the end users to develop applications. Cisco’s Devnet and HP’s SDN Devcenter are examples of these kinds of initiatives. I see a resemblance of this to mobile application development where Android and IoS vendors are allowing third parties to develop applications for their operating systems. For the networking application initiative to get same kind of popularity as mobile applications, there is still a long way to go. There is a need to have wider penetration of the Openflow switches and there is also a need for more compelling applications.
  • As explained in 1 of my previous blog on the models for White box switches, not all applications can fit into this Openflow + SDN controller model. Openflow does not scale well in many scenarios. There are many debates around the centralized controller and how it can provide scalability and resiliency. There will be network applications that scale only with a more distributed model and having the control and data plane closely integrated.


Pictures used in the above blog are from the references.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s