News Stay informed about the latest enterprise technology news and product updates.

Q&A: Rajesh Rajamani talks SDN challenges in production networks

Rajesh Rajamani discusses SDN challenges associated with production networks from a network service provider and enterprise network perspective.

Provider interest in SDN is growing -- in fact, 80% of global telecommunications service providers are including OpenFlow in their purchase considerations, according to results from Infonetics Research's Global Service Provider Survey. But as both network service providers and enterprises explore SDN, a number of implementation challenges have come to light that could prevent them from doing so.

We asked Rajesh Rajamani, senior product manager at Spirent Communications Inc., a UK-based telecommunications testing company, to break down some of the key benefits and challenges of moving SDN into production networks for both network service providers and enterprises.

What are some key challenges of moving SDN into production networks from a network service provider [NSP] perspective?

SDN is very promising because it has the potential to make networks cheaper to build and manage on an ongoing basis.

Rajesh Rajamani,
senior product manager, Spirent Communications Inc.

Rajesh Rajamani: Anytime you have a disruption of a complex network, you have challenges. SDN offers the promise of lowered Opex, lowered Capex, and faster provisioning of new, intuitive services. However, it's a fundamental disruptor for an established network because of the new technologies, protocols and network elements it introduces. SDN also presents many operational challenges, such as training existing employees [who are] well-versed in traditional network management and provisioning mechanisms.

The bottom line for NSPs considering SDN is to remember that end users do not care about the technical specifics of the network. They want streamed video to look crisp and uninterrupted, they want their phone calls to not be dropped, and they don't care how those things are accomplished. Operators need to ensure that on Day 1 of implementing SDN, the quality of experience meets acceptable standards (for example, latency, jitter, video quality and packet drops all fall within acceptable limits.) This is where performance testing is critical, and where SDN vendors need to prove they can support the overall QoE [Quality of Experience] and QoS [Quality of Service] at maximum scale.

Do enterprises face SDN challenges that differ from challenges you described for service providers?

Rajamani: Almost all enterprises have existing traditional networks that have been operating for many years. [and] their customers depend heavily on [these networks]. The core challenge is how enterprises go about upgrading their legacy networks to SDN. In most cases, the legacy systems, while being expensive to maintain, are not broken. It's a major challenge for enterprises to justify upgrading their still-functioning networks, so there are business-level challenges that must be considered to weigh the risks and potential rewards.

On the operational side, the way companies support their networks will fundamentally change. In some cases, OpenFlow-based switches and controllers will replace traditional routers and switches. In other cases, modern cloud management systems (e.g., Open Stack) will be deployed and used to orchestrate new and complex virtual environments on servers in data centers. Almost certainly, enterprises and service providers can expect to deal with new, independent software vendors to quickly bring to market revenue-generating services.

Testing of SDN-enabled switches, servers and routers is critical before deployment and needs to be done at the individual device level, the network performance level and [through] validation testing at the end-to-end platform level, where you test the entire platform, including any cloud management interfaces. For all three of these testing phases, you want to test the basic function and test at scale to the highest number of ports, control plane sessions and line-rate traffic.

Despite these challenges, what are the benefits of implementing SDN into production networks from both an NSP and enterprise perspective? Why is interest in SDN growing from both groups?

Rajamani: SDN is very promising because it has the potential to make networks cheaper to build and manage on an ongoing basis. Operators expect SDN to [allow them to create] new revenue-generating services faster. Gradually, we can expect standard, off-the-shelf servers and switches to replace expensive, purpose-built routers and switches. Simultaneously, we can expect many new software vendors [to be] developing new services that can be quickly deployed in virtual environments.

SDN and the service provider

Basic SDN for providers

Five SDN use cases

Why are providers looking to SDN?

For example, operators wanting to deploy new services in a new country can do so quickly by installing massive, off-the-shelf servers and switches, and then installing hundreds of network functions and services as virtual machines in a matter of days instead of weeks, without having to deal with shipping and customs hassles. Operators can scale as they go, depending on market adoption of their services.

Most importantly, SDN enables operators and cloud service providers to mine intelligence from the networks in real time and use the intelligence to dynamically reprogram the networks -- thereby providing value-added services to customers and increasing their revenue opportunities. An example is dynamic re-routing of data for important customers depending on traffic congestion in parts of the network and applying services such as firewall or intrusion detection services depending on threats in the network.

What other significant SDN challenges could emerge as the technology continues to grow and use cases come to light? In the same way, could growth in certain areas aid with current challenges production networks are seeing?

Rajamani: It is rare to be presented with a deployment scenario such as SDN, where you can build things from scratch. The downside to such a situation is that the risk of disruption is large. There is a lot of installed investment in hardware, especially for big operators, and introducing a new technology is a challenge. NSPs and enterprises need to judge SDN as they would any other promising technology -- specifically whether to pull the trigger on implementation immediately or wait to see what is coming on the horizon.

Another broader challenge comes from an SDN's ability to allow you to offer more services. Why is this challenging? Because as a business, you need to find ways to monetize these services. The SDN opens up many capabilities, but it's still up to you to determine what you do with them and how you ensure profitability.

Dig Deeper on Cloud Networking

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.