There was a new cult this year at OFC, the ‘Cult of Open.’ It formed from rising sentiment that the lack of interoperability of vendors in optical networks is a major problem and that the industry needs to embrace hardware dis-aggregation.
This is not a bad cult – it ignited a healthy debate that illuminates both the positive and negative aspects of open networks and closed single-vendor networks. But some members of the Cult of Open overlook an inconvenient truth – closed networks from one vertically integrated vendor can vastly outperform open networks in some cases. Cignal AI’s key takeaways concerning open optical networks and hardware dis-aggregation include:
- Facebook’s Voyager Platform
- Open ROADM and Line Systems
- Benefits of Vertical Integration
Open vs. Closed Optical Networks
Historically, the optical market was vertically integrated; optical hardware and software were deployed only from a single vendor. But now, network operators have a multitude of options when architecting and deploying networks. Operators can use a line system from one vendor and choose transponders and chassis from many other vendors. They can operate multiple network management and control planes in parallel, or find ways to unify them using either in-house or outside vendor software. Finally, if a critical component isn’t available, one can write the code or design a white box.
Hardware disaggregation allows for a “mix and match” that creates the best solutions, but the network operator must be willing to absorb the inherent integration and software complexity created. It was evident at OFC this year that incumbent equipment vendors are carefully crafting their products for both the open and vertically integrated models.
In the open model, vendors design equipment and software that is more open and independently controlled. This enables hardware disaggregation by operators and greater customization of their networks. Examples include AT&T’s CORD initiative and open ROADM, or the work being done at the Telecom Infra Project. Business models must be adjusted, such as pricing optical line systems higher than historical levels, because the sale of bundled higher margin transponders is no longer a given.
The traditional, vertical model is different. To offset the loss of customization, vendors emphasize their benefit to the customer through innovative pricing techniques, the reduction of unneeded margin in network design, or by offering hardware and software packages that provide the best cost and performance when used exclusively.
- A ‘Cult of Open’ is overtaking optical hardware marketing and conference discussions. The general perception is that open networks are ‘good’ and closed are ‘bad.’ Many vendors are putting forward solutions for both options, and some of the closed solutions are compelling too. A more measured view of the market is needed.
- The management and control of these open systems are, by far, the most challenging problem that must be solved. These systems were due for a transformative upgrade anyway, and as part of this process, the requirement to support multi-vendor systems is included. We can’t help but wonder if open systems are just a result of the transformation of network management and control, rather than a catalyst.
Facebook’s Voyager Platform
Voyager was a lightning rod in debates about open and white box optical networks at OFC. Facebook and ADVA are advocates and highlighted that numerous proof of concepts (POCs) and use cases were underway in customer labs. Detractors, who would prefer no attribution, characterized Voyager as “a solution in search of a problem to solve”, and derided it as a physical manifestation of the (perceived) disdain hyperscale vendors have for conventional optical suppliers.
Facebook designed Voyager’s first revision by using off-the-shelf components from Acacia and Broadcom. The hardware resembles a Compact DCI chassis and provides 800G of coherent transport paired with 1.2T of client interfaces, joined with an Ethernet switch fabric. Voyager allows the market to modify its design just as one might modify open source software.
ADVA offers Voyager for sale in conjunction with its OLS and network software. The company announced that nine network operators were currently trialing the platform. We spoke with one to get some first-hand impressions, and found a lack of enthusiasm for the product itself. The trial experience, however, was deemed enlightening and educational, with the operator realizing how ultimately dependent it was on the tightly integrated systems of traditional hardware vendors and how little the operator understood when trying to close optical links and make them work reliably. Using Voyager was a raw experience, something like building a PC from scratch as opposed to buying one from Dell preloaded with Microsoft Windows. When asked about Voyager’s cost, the operator could not answer; pricing had never been discussed.
In the opinion of this network operator, Voyager is not a viable long-term option in its current form, unless a customer places value on the deep customization it offers. This degree of customization may be a high priority for Facebook, but from the operator’s point of view, the challenge of making Voyager operational in their network outweighs any benefits.
If Voyager is to continue to gain traction in the market, it will need to win advocates beyond ADVA and Facebook, and it will be interesting to hear from future users about their motivation for using it. The reaction of most of the OFC attendees was that Voyager did not meaningfully address any deep industry problem, and the actions taken by incumbent vendors to build open Compact DCI chassis negate whatever threat it could pose to the status quo. If it is to succeed, Voyager must evolve into something more extraordinary.
We welcome critical feedback on this issue as it is top of mind for many of our clients; consider these as strong opinions, but weakly held.
- We simply don’t see what problem Voyager solves. Equipment vendors are already building systems that meet the exact needs of cloud and colo companies.
- Operator feedback about Voyager would indicate it is not ready for prime time from a systems integration standpoint.
Open ROADM and Line Systems
The OLS is the first step in implementing a more open model for optical transport. Network operators in the submarine optical market pioneered this concept by connecting a new transponder vendor to both ends of existing undersea cables to upgrade them from 10G to 40G or 100G. Ciena and Infinera both entered the submarine market this way despite owning no boats and having no undersea expertise.
This concept is currently being extended to terrestrial networks, where existing amplifiers and ROADMs from one vendor can work with several different transponder systems. Vendors are opening the software interfaces to their amplifiers and ROADMs, allowing for agnostic external control and coordination.
These days, most equipment vendors now offer open line systems and are working with initiatives such as Open ROADM. ADVA and Coriant announced such products, mentioned earlier. Fujitsu, a key AT&T partner, announced hardware designed to better control a mix of equipment, including Open ROADMs deployed within its network. Other companies, like Infinera, which sold substantial amounts of equipment early on to the Cloud and Colo community, embraced the open concepts these operators demanded. Now they have mature products to sell.
Led by vendors such as AT&T, the Open ROADM initiative details specifications and control interfaces which allow the construction and deployment of what is a “generic” optical switch. This approach lets AT&T, which is deploying Open ROADM, to mix and match transponder and ROADM suppliers within a single network.
An advantage of this approach is that it eliminates the long-standing practice of large carriers isolating equipment suppliers to specific domains – the practice of only deploying one vendor in a specific city or zone to guarantee interoperability and consistent operation. As reported by Fierce Telecom, Kathy Tse, Director of Photonic Technology Planning for AT&T, noted: “You’re able to mix and match different ROADMs and transponders, and it’s very critical to have the transponders inter-operate because you’re talking about a flexible network.”
The flexibility to select different vendors and cherry pick the best technology is one of the biggest advantages of open networks, as well as the negotiating leverage it gives operators over the supply chain. It also frees operators from the monolithic release cycles of vendors, in which advances can be slow to trickle out.
- Open line systems are the low hanging fruit of hardware dis-aggregation. There is a consensus that at least having this capability makes sense.
- Vendors are excited about open line systems; in the past sales of common equipment were an up front low-margin business required to secure higher margin business later. We’ve yet to speak with a vendor that doesn’t like moving to open line systems.
Benefits of Vertical Integration
Implementing and deploying open networks with many different vendors does have a downside, and it is margin stacking. By definition, interoperability requires all vendors to drop performance to the lowest common denominator to work together. The worst case operational specifications for each vendor must be assumed when integrating different components and systems. As a result, it becomes difficult to fine-tune for maximum performance.
Several vendors, all of whom are also making a big push to open their networks, also announced hardware and software solutions that seek to optimize the performance of these networks. And the performance of these systems is optimal when the end-to-end solution can be assumed to be vertically integrated from a single vendor.
Infinera is one company that has a history of making vertical integration a priority. It used OFC to announce its Instant Networks, a major expansion of its Instant Bandwidth feature, which allowed operators to dynamically turn up extra capacity in Infinera hardware on a link-by-link basis. This process involves licensing unused wavelengths inside the highly integrated photonic circuits the company uses as building blocks for its network.
The result is virtual capacity that can be enabled and moved flexibly, particularly when paired with automated capacity engineering. This type of approach would simply not be possible in an open network and relies on the tight vertical integration of custom Infinera hardware, software, and commercial agreements. One final observation – by decoupling the purchase of capacity from the purchase of physical hardware, operators become less concerned or aware of the underlying technology. A service provider may not care if traffic is carried on the latest hardware or old hardware – the customer is just paying for capacity. Instant Networks has the potential to be both more efficient and flexible than an open network – but one must hand the keys to the kingdom to Infinera.
A week before OFC, Ciena unveiled Liquid Spectrum, a tool for automating planning in optical networks. It uses the deep performance monitoring feature sets built into Ciena hardware such as WaveLogic Ai. It extracts more capacity from the network by optimizing coherent modulation and wavelength spacing routing, and by removing extra margin from the network when it isn’t needed. This is accomplished through additional software applications built into Ciena’s Blue Planet software, which can operate in a multi-vendor environment. However, Liquid Spectrum works best if it is operating on the highly instrumented Ciena photonic layer. Coriant made a very similar announcement with its Aware technology presentations at OFC.
The simultaneous advances in open optical networking and tighter hardware/software vertical integration will continue to provide better and more competitive options.
- Ciena and Infinera are offering an alternative to the ‘Cult of Open.’ Ciena has invested in the Liquid Spectrum approach for years – we’ve seen the groundwork laid during annual visits to their labs in Ottawa. This is not a paper press release and provides yet another way for Blue Planet to take root in their customer’s network.
- Infinera is taking its Instant Bandwidth concept network wide, leveraging a unique aspect of its hardware – latent capacity the customer has installed but not paid for. There are a lot of ways Infinera can take this business model, and it isn’t something others can duplicate.
- It is refreshing to see equipment vendors simultaneously offering open networks and hardware dis-aggregation, while also tuning their solutions to be high performance. It will be interesting to see which customers choose open vs. closed when faced with two optimized choices.