![Industry Perspectives Industry Perspectives](https://eu-images.contentstack.com/v3/assets/blt8eb3cdfc1fce5194/blt79a1824cfad941c4/664ba129be90a086233eb8e0/Industry_Perspectives_DCK_light.png?width=700&auto=webp&quality=80&disable=upscale)
Insight and analysis on the data center space from industry thought leaders.
Sticking Point: SDN Management Challenges
Achieving the many promises of software defined networking (SDN) needs to happen in evolutionary – not revolutionary – steps. One evolution that needs to happen is the adaptation of network management processes and tools that can keep up in a programmable world, writes Cengiz Alaettinoglu of Packet Design. Management processes and tools always seem to lag behind, but they are especially critical in SDN.
February 11, 2014
Cengiz Alaettinoglu is CTO, Packet Design, where he provides the technical direction for the company’s portfolio of route analytics products as well as the prototype of an SDN management tool.
CengizAlaettinoglu-tn
CENGIZ ALAETTINOGLUPacket Design
Achieving the many promises of software defined networking (SDN) needs to happen in evolutionary – not revolutionary – steps. One evolution that needs to happen is the adaptation of network management processes and tools that can keep up in a programmable world.
Management processes and tools always seem to lag behind, but they are especially critical in SDN. This is because the human operator’s visibility and control is curtailed, management tools cannot match the automation SDN brings, and there can be great variability of traffic demand. And as with any new technology, SDN is being deployed in mixed environments, making management even more difficult.
Also, deploying SDN across multiple data centers adds another layer of complexity. SDN is an island within a single data center, but across the WAN where resources such as bandwidth become scarce, management becomes much more complicated.
All this begs the question: How can you adequately manage an SDN environment under these conditions? Whether the network is programmed or configured (or a combination), it’s not impervious to faults, including link or node failures. It’s difficult to plan for new applications and services if you do not understand the current and historical traffic load to ascertain how changes will impact existing applications. For troubleshooting, you still need to be able to compare and contrast the current network state to a baseline and find the root cause of problems quickly.
In addition, SDN introduces additional risks, including failure of the controller itself and the possibility of multiple controllers issuing contradictory instructions to forwarders.
The bottom line with SDN is that you still need to be able to manage the network according to time-tested management practices. You need to audit the network to make sure it’s healthy, including the integrity of flow paths, for instance. You need to understand what the footprint is now vs. the programmatic changes that the SDN controller will request, to ensure that the required resources are available.
Network Virtualization
One clear example of the need for SDN management is in network virtualization. In an SDN environment, you must be able to simulate moving a virtual machine, whether it is inside a VPN or not, from one location or data center to another, as well as all the flows originating from it. Also, you need to visualize and analyze the impact the move may have on any of the new network paths (such as congestion) and if so, what other services are effected. The ability to model modifications to the network and to flow records in real time is critical.
Also, if an application is going to run for a long time, it is important to understand historical traffic loads to predict loads in the future. For example, what if the SDN controller makes a request in a trading network a few seconds before financial markets open? Traffic loads will change dramatically once market data starts flowing. How do you know if the request will negatively impact the trading application or not? Past traffic volumes can be used to predict future traffic profiles, and these profiles can determine whether the application should be permitted to run or not.
Automation Needed
What’s needed is to translate today’s tried and true management techniques – including some of the manual functions done today such as planning – to ensure the network can handle what’s being requested. We need to apply automation to all responsible management best practices, including controlled configuration and provisioning, sustained monitoring for availability and performance, efficient troubleshooting, and effective security and policy governance.
The management vendors are lagging behind right now, but network managers and directors need all the help they can get as SDN adoption becomes more widespread. SDN will not make these network professionals obsolete. In fact, their knowledge and experience is more crucial and valuable then ever to ensuring optimal network performance. But they need the right management tools to be able to visualize and analyze what’s happening in the network. Only then will SDN fulfill its promise.
Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.
About the Author
You May Also Like