Amie Cox, PhD is Global ICP Sales Leader for VIAVI Solutions.
For today’s internet content providers (ICPs), the disaggregation gamble has paid off. Recent efforts to disaggregate hardware from software in the data center have not only radically changed the data center model, they have created unprecedented innovation. This focus on open computing has contributed to lower equipment costs, reduced power consumption and greater flexibility in systems operations. So far, so good.
In fact, the move away from proprietary hardware to less expensive servers and switches has been so successful, the industry is now looking to apply this white box approach to data center interconnect (DCI) networks. Yet, disaggregation of the Dense Wavelength Division Multiplexing (DWDM) optical networks that connect data centers will bring a whole new set of challenges that are likely to catch many ICPs off guard.
Unlike the first wave of disaggregation, which took place in a controlled environment, a white box approach to the optical network will have implications extending well beyond the confines of the data center. What would this mean for ICPs, their workloads, and the performance and interoperability of their networks?
Open Computing: The Next Chapter
The Telecom Infra Project (TIP) aims to create greater innovation and flexibility through disaggregation of the traditional network deployment approach. However, not only is this industry initiative addressing similar hardware issues as the previous Open Compute Project, but they are also looking to disaggregate terminal equipment from the optical line systems.
With a traditional network, optical line system components are designed and optimized by a vendor based on a set of proprietary standards. While the move away from proprietary equipment would be a win for disaggregation efforts, this means the ICP will no longer be able to rely on a single vendor to manage the design and validation of new network systems. As a result, ICPs will be on their own to tackle challenges that most of them have never had to deal with in the past.
Today, most ICPs trust their system vendor to validate the performance and interoperability of a new optical network. With disaggregation, however, ICPs will need to do all of the stress and performance testing themselves for every aspect of the network to ensure its end-to-end performance and reliability. Just stringing together a collection of white boxes and hoping for the best is not an option. In order to succeed with a disaggregated DCI network, it will be essential for ICPs to address a number of potential problems throughout the network lifecycle: during installation and turn-up, while running live traffic, and when troubleshooting.
Verify System Compatibility
When installing a new network based on disaggregated infrastructure, the first step is to test for compatibility. An end-to-end connectivity test can be completed using an agreed set of parameters to determine if the line system components are working together. If the test fails, the ICP will need to isolate and resolve potential causes – not necessarily an easy task when dealing with a white box system. Is the problem due to incorrect buffer settings, or a misconfigured switch; is a new component out of specification? Perhaps the bigger question for the industry to address is, ‘who is responsible if the components are not compatible?’
After compatibility has been validated, the next step is to test the physical layer of the network, starting with an inspection of the fiber to make sure the connections are clean. To ensure the integrity of the fiber, ICPs will need to test for insertion loss and Polarization Mode Dispersion (PMD) while also deploying an Optical Time Domain Reflectometer (OTDR) for various tests. The final phase of installation will include testing for throughput to ensure the new network will be capable of handling the highest possible levels of traffic.
The ICP is now ready to turn-up the system and introduce live traffic. Installation testing may sound tedious, but it’s much easier to fix potential problems before traffic is introduced. Plus, results of the tests completed so far will create an important baseline of what performance should be for the network, facilitating future troubleshooting.
Monitor Live Traffic
Once live traffic is introduced to the network, the ICP should have a system in place to monitor performance. Since the primary impetus behind disaggregating the DCI network is to increase flexibility and speed, it’s necessary to monitor traffic to determine if performance goals are being met. And as bandwidth demand continues to grow, the highest possible level of performance and throughput will be required to meet demands on the network.
As an added benefit, having a monitoring solution in place enables ICPs to spot transport and service performance degradation problems before they become an issue for end users. Armed with baseline performance data from the install phase, an ICP can then compare current performance levels to what they should be in that particular line system.
Being able to pinpoint a problem, and where it is located, before it affects live traffic is critical to helping ICPs manage network performance issues more proactively. Of course, when an issue is not caught preemptively, the ICP is responsible for troubleshooting and fixing the situation as quickly as possible.
The same tests used during network installation and turn-up can be used for troubleshooting. But it’s important to be thorough, since one large problem often can mask other issues. Therefore, after the issue is identified and resolved, the ICP should again compare with their baseline data to ensure the network has returned to optimal performance.
Solid Foundation for the Future
While these challenges represent a new normal for ICPs, that doesn’t necessarily mean they have to put the brakes on optical network disaggregation. The innovation and collaboration achieved to date within data centers has been impressive, and pursuing a similar approach in DCI networks could create much more flexibility and speed. Yet the next phase of disaggregation will come with added responsibility to make sure network components work together, and vigilance to ensure optimum performance throughout the network lifecycle.
Opinions expressed in the article above do not necessarily reflect the opinions of Data Center Knowledge and Informa.