Whether it’s the proliferation of cloud applications or growing cybersecurity demands, data centers are being forced to change faster than ever. Pushing 1G or even 10G architectures to their limits won’t work moving forward.
That’s why SDN is slowly starting to become a commercial reality (albeit slowly). We’ve looked ahead to where SDN will be in 2020, but you could argue that that’s been SDN’s problem since 2013—we keep looking ahead and commercialization continues to be delayed.
As you’re scaling a 40G or 100G network, you need to know if SDN is available right now. Moving into 2017, is SDN really feasible for you?
Current Adopters Face Challenges with SDN
Even though SDN hype has dwindled, companies are investing at a strong pace. While many of these companies are in testing phases, commercial deployments are popping up.
According to a recent eWeek survey, companies that are using SDN to some extent are seeing improved network performance, increased productivity, and simplified network operations—so you know SDN isn’t out of reach. However, these are the challenges that you can expect:
- Cost Savings: Oddly enough, cost savings was both the top benefit and top challenge found in the eWeek survey. The problem is that your cost savings will initially depend on the investment you have in traditional switching/routing technology. Depending on how smooth integration goes, your cost savings could be a perceived challenge.
- Security: Centralized control is great, but it might mark open season for attackers. Securing SDN requires a different approach than traditional architectures and it seems that’s taking some getting used to for some adopters.
- Integration and Interoperability: It’s unlikely you can support a wholesale change in your data center—implementation will come in pieces and companies are having a hard time integrating deployments effectively.
The integration and interoperability challenges are one of the biggest problems that have kept SDN from realizing its potential. If you’re looking to leverage SDN in 2017, you have to understand the issue.
Protocol Sprawl Leading to New SDN Expectations
The original promise of SDN—the one that had people so excited for a new generation of networking in 2013—was for centralized control of traffic via a programmatic interface. This would be the SDN controller that managed southbound interaction.
These SDN controllers manifest in open protocols that companies integrate into their deployments. But this is where the interoperability issues emerge—there are just too many protocols and no set standard as of yet:
- OpenFlow: The leading choice to become the standard SDN controller, despite a host of challenges that have led other initiatives to become popular.
- NETCONF/YANG: Based on the remote procedure control, this was a key competitor for OpenFlow adoption. However, OpenFlow has since embraced NETCONF.
- ECOMP: AT&T’s open source project that came as a result of OpenFlow stagnation.
- Cisco OpFlex: A protocol that embraced SDN’s flaws and stopped focusing on openness as a strict requirement.
These are just a few of the choices competing for SDN deployment attention. While some believe 2017 will be the year that a standard is agreed upon, it seems more likely that 2017 deployments will accept a lower expectation from 2017. Rather than a strict programmatic interface, centralized traffic management may suffice.
If you want to enjoy the benefits of SDN in 2017, centralized management may be your best bet rather than hoping a protocol like OpenFlow will integrate perfectly into your data center. The key is to implement this centralized management with the right architecture of network TAPs and packet brokers.
If you want to learn more about adopting SDN in today’s 40G and 100G data centers, download our free white paper, Architecting Data Centers for SDN and NFV.