Software-defined networking (SDN) can be confusing if you don’t know what it is. Virtualization has revolutionized storage. Data was stored on individual servers’ disks, and only the storage controller of that server could access it. If the server fails to boot, you can’t access the data or initialize the storage.
SANs and storage network allow us to separate the storage from individual servers. Although the SAN is limited in intelligence, it still relies on the OSes of servers to interpret ACLs and partitions. The storage is centralized, which means that the servers are not at risk of being harmed.
SDN is similar to networking. The intelligence that determines how to handle, shape and route traffic (usually called control plane) is removed form the individual pieces of network hardware. There are also management planes that allow you to change configurations, update firmware, and so on. The central controller now manages everything, with an end-to-end view across the entire network. The data plane is stored in the box, which sends and receives the 1s.
The advantages of an Intelligent Network
Learn how to become a security expert with SPOTO’s Cybersecurity Training
Start trainingThe controller is able to make better routing decisions. The controller can see all information, unlike individual routers that communicate with each other to determine their position in the network. Not only are shorter routes advantageous, but traffic can also be automatically rerouted in case of an outage.
It is also much easier to make configuration changes. Do you need to add a VLAN/ACL to 10, 20, or 2,000 switches? It will be painful to log into each one and run the required commands. Although scripts can be written to automate the process, they must still be maintained and adjusted as the network changes. SDN allows you to log in to your controller, make changes, and push it out.
It is easy to see the cost savings. Config changes can save you a lot of time and administrative overhead. Misconfigurations can cause outages and other problems. You can no longer rely on vendor lock-in and pay the associated costs.
Both successes and failures
Google is a great example of how to leverage SDN in real life. Google has one of the most powerful networks in the world. They have created their own SDN network stack called Andromeda, to manage their network. Google describes it as the orchestration point for provisioning and configuring virtual networks and packet processing. Imagine them manually changing their vast global network. You’ll be amazed at how SDN works.
Google is not perfect, however cool it may sound. Google Cloud Compute experienced an 18-minute outage on April 16, 2016, but the post-mortem report sheds light on the day to day operations of Google’s SDN setup.
The software bug that removed all advertised IP blocks for Google Cloud Compute caused the outage. This effectively stopped all inbound traffic. We can see SDN features throughout their setup. Software automatically pushes config changes out across the global network. Failsafes to verify changes as they are rolled-out across sites. Traffic is automatically routed to online sites and reverted to the original configuration.
The outage was caused by problems with SDN management software, but it also allowed for quick remediation. We learned lessons, the bugs were fixed, and we are certain that no one at Google said “Maybe we should abandon this SDN stuff” and instead go back to manually configuring routers across six continents.
If it’s good enough for G