“As Mike Tyson said, ‘Everybody has got a plan until they get a punch in the face.'”Rob Keys, Senior Director Optical Transmission, Ciena at OFC 2021
On our second, closer viewing of the Market Watch panel, “Evolution to Coherent WDM Integration in Routers,” while not preoccupied in getting an analysis out in real time, fibeReality now believes that Keys was kind of acting at least a bit like a hatchet-man and that there was at a minimum, a veiled attack at Microsoft, by far the biggest promoter of 400-gig ZR. (This is one concept from Microsoft that we like, as we are extremely bullish on pluggables, given the growing amount of pliability with length and capacity in the same slot.) Keys’ speech seemed to be more than advancement of a full turnkey prerogative in its battle to fend off the looming threat from 400ZR modules and other future pluggables in the hyperscale space. While again, given its network requirements, we certainly view the press by Microsoft for 400ZR to be logical, there does appears to be a slowly moving, outward resistance to some of its potentially damaging tactics and expectations in the the optical space as a whole. Surely, Arista Networks has been unusually outspoken against co-packaged optics, albeit after its level of business with Microsoft declined significantly. We have also suggested that Google’s endorsement of the OSFP-XD concept is a sign that it will no longer publicly be silent, to a certain level indirectly, about its dislike for CPO, and perhaps other ideas pushed by Microsoft, such as through COBO, like conceivably, expanded beam optics, in which resources are compelled to be used by suppliers in an ineffective way. Even Facebook has moved in its own direction on CPO, as we pointed out at an EPIC conference at ECOC, earlier this year. There are other examples of suppliers not totally supporting the webscaler’s plans anymore, such as Intel, which arguably has been Microsoft’s biggest ally in the past on optics direction. Otherwise, previous attempts by Ciena with the investment community to allay fears, such at its “Coherent Pluggables Chalk Talk,” did not appear to come across nearly as convincing. Once there is a takeoff in the 400ZR space, the users will be appreciably empowered and some of supplier’s negotiating ability will be weakened. Yet, we remain confident that Ciena will retain its dominant international position for a very considerable amount of time.
In elaborating on the “plan,” or “blueprint,” it involves the “400ZR IA greenfield view of the world. If you have a legacy network, you are quickly going to get a punch in the face because you will probably find that it doesn’t meet exactly those standards that’s been defined.” Keys mentions ROADMs, legacy transponders, the different spectral occupancy, and the disparate bauds. (As we note below, an obsession to a fault with standards could be an apt description of Microsoft, and then there has also been the unanticipated degree of difficulty to be overcome with interoperability with diverse vendors’ 400ZR gear.)
“The challenge is how do you deploy these new pluggables on that greenfield site,” according to Keys. “One of the things we find is when we are having these conversations about the performance of 200-, 400-, 800-gig, they are all running at different bauds. It’s a very difficult conversation when we start talking about ROSNR — it scales with baud, and you always have to be doing logarithmic calculations in your head.”
Keys then states that it moving right to the receive SNR, the discussions and the comparisons become extremely simple with improved performance. He also advertises the employment in the metro of the “Ciena implementation of the standard, where you don’t have to worry…about interoperability. In exploit[ing] some of the more proprietary modes, run a slightly higher baud, [moving to] better FEC schemes [as well as] better constellation shaping, better filtering shaping, [using 1×9 WSS gear], we can cover about 96% of the desired A- to-Z demands of the -ZR pluggable modem with the colored mux [as opposed to achieving approximately] 60% with the [device] compliant with the OIF ZR MSA implementation agreement — [just] 22% with a colorless mux.” (David Boertjes, Director Advanced Photonic Networks R&D at Ciena also emphasized an SNR-centric transport layer during his session.)
“The takeaway here is — what’s important to you? Is it just interoperability or high performance?”
This statement by Keys could be easily taken as a subtle shot against Microsoft, which is willing to take on generally, for example, more cost, such as additional fibers, in the name of standardization. The prominent example has been the 100GbE PAM4 QSFP28.
“It’s a bit like filling up a car up to go on a family vacation and when you drive off the driveway, you lost about 70 percent of your gas. That’s not a great scenario.”
The analogy by Keys referred to a network constructed for 100G, with a colorless multiplex, in eventually moving to 800ZR, whereby “the incremental noise contribution of the [mux] takes up about 70% of modem budget because of noise funneling, etc. — and supports just two spans. Colorless muxes are fantastic from an operations point of view, [but] as you start to move to higher bauds, high throughputs, the noise contribution can be significant.” Keys advocates plugging the 800-gig pluggable right into the WSS port with a colorless mux that is also filtered, where “only 20 percent of the modem SNR budget is consumed by the mux and up to seven [transmission] spans get to be covered [in this particular illustrated case, as opposed to only two spans otherwise].” (Although with intra-data center applications, Microsoft plans to go directly to 1.6 Tb, we would be extremely surprised if it does likewise on the line side, and totally avoids 800ZR.)
In fairness, Keys certainly does not discount the feasibility of cases in which there will be total openness with routers, ROADMs and pluggables over an OLS, and asserts that Ciena knows how to accomplish that task. He requests help from the user community in defining interfaces, in standardizing the APIs, workflow, etc. Obviously, he has to say that from a political expediency standpoint, but greater interaction provides the supplier an opportunity to tout the advantages of a fuller optical networking solution.
Regardless of specific intent, this presentation was more powerful than the “chalk talk” mentioned above, partly because the onus in making pluggables work quite effectively is proven to not be all on Ciena or the other suppliers. The webinar in April, followed by the Q&A session with the financial analysts over 400ZR modules, was not as credible for the following nine reasons:
- It supported the wrongheaded notion that any kind of technical differentiation/innovation at hyperscalers could ever result in anything resembling a premium for a vendor.
- There was the submission of the 20 to 30 percent price erosion that the industry has been dealing with for many years (our intelligence has indicated an extremely precipitous decline in the expected cost even before initial shipments).
- The contention that only a small handful of hyperscalers desire the 400ZR is not forceful enough in making the point above that we are essentially only talking about two major webscalers (this point had to be reinforced with the help of an analyst), whereas the lack of specificity could falsely imply something much greater, such as the previous, universal move to OLS by the cloud companies. Clearly, it is only Amazon that has a similar infrastructure to Microsoft, whereas Google will have a relatively limited requirement, and Facebook has no metro assets.
- The inherent contradiction of using “a set of contract manufacturers” for the 400ZR (undoubtedly to get the cost hit as low as possible), while struggling to make the case that its high level of vertical integration will make a substantial difference.
- The rambling recitation of the various market segments, while not harping fully on the fact that the -ZR play is essentially campus-metro DCI in nature. Furthermore, making the connection with future coherent capabilities in the access, ignores the continuing reliance on PON in that segment, which is not a strength for Ciena.
- Making the case that the WaveLogic 5 Nano is the “culmination/next stage of its [business/technical] experience” — and saying that the “Nano variant is…equally impressive with the rest of its product line” all come across as pure hyperbole. The transponder business represented the lowest end of the food chain for Ciena to be ensured of avoiding a noticeable drop in its sales and to meaningfully take advantage of its strengths as a system house. There is no coincidence that Ciena stopped reporting its Waveserver revenue.
- The unrelenting message from Ciena that there is a merchant market goes beyond competitors not desiring to do business with such a competitor. In the past, Ciena itself placed barriers in the way. Then still bringing up in 2021 that licensing and the like was about making penetration into China, just leaves one speechless.
- While Ciena can feign being insulted by its reputation for only high-end coherent expertise, while at the lower-end, it has a great deal of knowhow, these discrete components were enablers, not what played an instrumental role in driving the company’s business forward.
- Bringing up “a lot of activity going on co-packaged optics” and how that could mean far fewer pluggables, is about throwing in everything but the kitchen sink.
To be sure, Ciena got a few licks in with positive statements at the “chalk” conference, such as the “400-gig plug does not simply plug and play into all network apps.” It was also contended that “in…other use cases, we actually still see this being very much a system player game, because there’s a lot of other complexities and capabilities that come into delivering the end-to-end solution to that customer base, even if part of that solution is going to be pluggables.” Still, the most revealing comment by McFeely happened to be — “we’re on a treadmill and we have to continue to win in this business.”
The problem for the system integrator is that the demand from the hyperscalers for devices is likely to continue to get bigger compared with the traditional service providers, which will not grow comparably in size. We fully expect some additional acquisitions by Ciena to make up for the loss of market share.
In our judgement, CTO, Steve Alexander, would have been much better in offering a cleaner and more direct explanation of the 400ZR situation. In contrast, Senior Vice President & General Manager, Networking Platforms, Scott McFeely, is too invested personally in the “Nortel” optical system evolution story, and is unwilling to acknowledge that going beyond transponders in a pizza box, will definitely be a bridge too far in many circumstances.
Ciena’s ability to thrive for an indefinite period of time is based on its positioning in the optical system marketplace from a total solution standpoint. Keys did a better job than McFeely in driving that messaging home because the former went on the offensive (he also had the advantage of being able to adequately plan in going second). Ciena has to be a participant (compared to about a year ago in our report on 400ZR, our latest intelligence points to the supplier having a product that is competitive with others) in this space to mitigate the amount of damage with a fuller offering and to avoid alienating the hyperscalers from future business, which will be more lucrative.