Open Optical Systems
Just Because You Can Doesn’t Mean You ShouldNetwork operators are intrigued by the concept open optical systems, whereby they can assemble an optical network by piecing together parts from various vendors. They are encouraged by concepts like disaggregated transponder boxes for DCI applications, an Open ROADM MSA, and the Telecom Infra Project Voyager white box initiative. While there is undoubtedly increasing interest in the open model, the question is how far can and should this model be pushed.
Let’s start this discussion with operator motivations. A recent IHS survey ranked the top three motivators for open optical systems as, (1) lower capex through better negotiating power, (2) less reliance on a single vendor, and (3) availability of SDN control. There can be no arguing with these desires for lower cost, more choice, and greater control. These motivations have always been a part of network operator decisions.
But can open optical deliver a solution to those problems? While vendor choice and granularity of control certainly is proportional to the degree of disaggregation (aligning with motivations #2 and #3), so is the amount of software an operator must develop for system integration and management, adding to complexity and total cost. And as for first cost (motivation #1) it is difficult to say whether assembling pieces individually is intrinsically less expensive than purchasing an integrated whole.
Let’s see how varying levels of disaggregation play out for building a moderately sized optical backbone network, of say 100 nodes.
At the lowest level (least amount) of disaggregation, a single vendor supplies the entire optical network. While by definition, this network depends on a single vendor, we can assume that it is fairly priced based on a competitive bid situation. Pricing includes ongoing maintenance, software upgrades, and support, all of which allow the vendor to offer lower initial equipment pricing due to the promise of future recurring revenue. The network operator doesn’t need to invest any effort into system integration, and all network management and control is available through an open NMS or SDN controller API, for higher level packet-optical services orchestration.
At a medium level of disaggregation, several vendors supply individual nodes of integrated network equipment (NEs). A likely scenario is that different geographic parts of the network are parceled out to different vendors. We can assume similar pricing to the first situation on a per-element basis, but with perhaps higher long term cost of ownership. While there may be more competition for each part of the network, there is less economy of scale and each vendor will need to make their profits on a smaller part of the whole with less opportunity to negotiate total ownership costs due to lack of secured recurring revenue. In this level of disaggregation, the operator must start investing in system integration, and developing management and control systems that aggregate open EMS or SDN NE-level APIs. This software development effort, while not insignificant, is probably within the scope of many operators, and can lead to customized and streamlined operations.
A point of caution here is that optical systems are fundamentally analog and usually incorporate proprietary transmission engineering. In order for transponders from different vendors to pass traffic, they typically must be set to some lowest common denominator for performance. Most optical vendors use advanced forward error correction (FEC) schemes that allow operation over longer distances and/or more ROADMs. But those advanced FECs do not interoperate. Only the lowest level “standard” FEC can be expected to allow interoperation, and even then many systems will not work well together. Interoperability of transponders is still not a given for the industry, especially at higher speeds.
Likewise, building networks with elements from multiple vendors will eliminate the possibility of using any advanced network monitoring, communications, or operations capabilities that have been developed by the vendors, at least network wide. Most optical vendors have a unique set of differentiating software features to make their equipment easier to use, assuming that their equipment is deployed throughout the network. In-band signaling for amplifier adjustments, performance insights into component health, automated responses to specific inputs, and other features can have significant benefits that may be elusive in an open systems model.
At the highest level of disaggregation, multiple vendors supply stand-alone subsystems like transponders, ROADMs, and amplifiers. Any CAPEX savings from full disaggregation would be hard to prove against the previous scenarios since each subsystem now needs to bear the overhead of physical packaging and common equipment (fans, operating system, etc.). Although some savings can perhaps be obtained through careful selection (for example, by using less expensive subsystems for parts of the network where lower performance can be tolerated), any complex network built from completely disaggregated subsystems with similar performance to a traditional integrated approach, would most likely have a higher total cost of ownership.
For this highest level of disaggregation, a network operator also must invest significantly in system integration and homegrown management and control systems. These must aggregate many hundreds of open subsystem APIs. This capability is probably beyond the scope any but the largest operators in terms of resource availability and expertise. This software complexity would be even more the case for subsystems disaggregated as white boxes, where APIs now become firmware system calls. Disaggregated networks using open subsystems are primarily being investigated at this time by large Web 2.0 companies with significant software resources to tap and relatively simple point-to-point networks to operate.
That said, there are certainly niches within even moderately sized optical networks where subsystem level disaggregation can make sense! One application is to support real-time data replication (aka synchronous backup), where a change to a database in a primary data center is not committed until it is duplicated in a secondary data center. Synchronous backup requires very high bandwidth and very low latency, typically over shorter distances (up to 80km) with no requirements for ROADMs or mid-span amplifiers. This simple high bandwidth point-to-point “patch cord extension” application could be ideal for disaggregated transponder boxes.
Another application may be to implement an optical switching hub for links coming and going from many nodes. Without a need to integrate with transponders this may be well suited for disaggregated ROADMs.
Putting this together, we can start to see a picture emerge of where open optical systems are likely headed over the next few years.
- The bulk of an optical backbone in moderately-sized (e.g. 100 node) networks will continue to be provided by a single vendor. However, this will be accompanied by an increased emphasis on open and powerful network control APIs for integration with service orchestration applications.
- Geographically bound or functional subsets of the optical backbone will start being awarded to second vendors. This solution provides benefits of more customized control over the network and ensures competitive pricing. This step will be undertaken by network operators with operations organizations that possess sufficient skillsets and resources to perform the necessary system and operations integration.
- Islands of disaggregated subsystems will be deployed for niche applications that deliver significant cost, performance, or control benefits, with relatively low system integration obstacles. Examples are point-to-point data center interconnection or ROADM hubs.