BLOG

Cisco Systems’ Optics Pastime

September, 2018

It all began for Cisco Systems last century during the bubble with its purchase of Cerent for a whopping amount of money, as it was a startup merely providing a basic SONET transport system. Although for about 15 years, Cisco made optical networking a prominent part of its corporate messaging, since then, it has been pretty much ignored at that high level in presentations, and any comments about the technology are confined to an occasional announcement, such as at OFC. Also, optics development is restricted to a relatively small, seemingly contented operation without too much pressure of having to make a significant impact on the bottom line, especially with the overall performance of the firm having improved in the last couple of quarters. While it had to be a minor relief to the optical division that the Service Provider (SP) segment finally had some revenue growth to report after over two years of decline, maintaining the number two position in North American market share in optics is probably just fine with the leadership for now. Although there will likely continue to be uncertainty about the SP business in the foreseeable future, which in the long-term, could even adversely affect the optics piece, fibeReality suspects that the company will hang in there indefinitely, particularly on targeted SP operating expenses, with the promise of 5G wireless having a substantially positive impact. We also believe that this hobby type of optical activity, which we estimate in the neighborhood of just two percent of total revenues, is sustainable enough to support any necessary R&D work, as part of its greater homegrown effort on MPUs to architect its switches properly in accommodating higher data rates, as well as in making minor refinements on existing transport solutions in the marketplace. One of the more interesting questions is whether the influence of the founder, Kalpendu (Kal) Shastri, of Lightwire, a Silicon Photonics (SiPh) supplier, which was acquired by Cisco, still remains at all at the company, despite him leaving the firm a while ago.

Cisco Systems claimed that the product from Lightwire “enabled us to deliver a terabit line card on a router in 2013.” On the other hand, the idea to use the Lightwire technology for a proprietary form factor, CPAK, was eventually abandoned by Cisco. We understand that Cisco has contended that it purchased the company more for long-term applications.

Shastri was one of the pioneers of electrical SerDes going back to the late 1990s. It appears that initial devices operated at 155-meg. These engineers were very clever in allowing them to last to the present.

Shastri has been in the camp of those advocates viewing the move to greater capacities as a dramatic packaging exercise. The expectation is that there will be the creation of a serious predicament in ultimately running out of electrical SerDes, and that the industry needs to be ready for this outcome. The folks with this perspective further admonish that as the electrical trace lengths continue to get smaller, or as the number of pins in a backplane become more taxing, copper will ultimately not be sustainable.

In other words, according to this point of view, there will eventually be no more serializing and deserializing, and the internal plumbing of how ASICs will talk to each other will be based on returning to the designing boards from about 30 years ago, in which these chips would be in a parallel arrangement. They would be similar to cell phones in which the chips are packaged on top of each other, and not spread among the boards anymore. We have recently discussed our thoughts on the high likelihood of pluggables not going away anytime soon, the challenges of chip-to-chip concepts, such as co-packaged optics, and that if they were to occur, the use of discrete optics would definitely be preferable over SiPh.

On the optical side, Cisco has not been prone to making major leaps technologically, as pointed out by its vehement resistance to OSFPs. The role of the interconnect, including the optical transceiver, is to as cost-effectively as possible get out of the way of the MPUs and the packet processors, allowing them to run as fast they can – it should not be the limiting factor, and that mentality initially led to the failure of OSFPs to become at least the de facto, standard form factor for 400-gig -- although not all large (see our comment here) users necessarily adhere to any kind of standardization. While the OSFP arguably had some merit in addressing some longer lengths, the density hit on the faceplate made it undesirable.

As always, fibeReality does not recommend any securities, and this writer does not invest in any companies being analyzed by us.

To follow us on our totally separate, quick update company blog, which is exclusively on fibeReality’s LinkedIn page, please click here.

[written by Mark Lutkowitz]

 

Share Article

Add Comment

 

Back

 

Read Comments