JPM Notes on Infinera plus more observations

From JP Morgan:

DCI Concerns Overdone, Infinite Capacity Engine Positive - ALERT

We attended Infinera’s investor presentation at OFC and came away unconcerned about recent Microsoft DCI announcements. We believe that direct detection optics remain useable in only shorter reach (<20km) scenarios while coherent solutions are required for longer haul applications. We also see the potential rollout of products around the “Infinite Capacity Engine” as a possible driver of better revenue growth later this year and in 2017. After the selloff yesterday, it presents investors with a buying opportunity for a high quality optical company heading into a significant product cycle and strong continued 100G and DCI business.

*** DCI concerns overdone. We believe that concerns over Microsoft adopting direct detection DCI technology have been overdone for Infinera. After talking with both Infinera management and multiple industry contacts at OFC we believe the range of direct detection technology remains too short for most long haul DCI applications. We suspect that MSFT may intend to deploy the tech on very large Data Center campuses rather than over longer distance. We note that Inphi themselves is hosting a discussion of direct detection at “experimental” distances of 40km.

*** DCI L3 a possibility. When asked about the threat of direct detection optics on Infinera’s DCI business Dave Welch pointed out that it is much easier for a company with a superior power and density solution in optical to reverse into packet networking than it is for commodity [emphasis added] based packet networking companies to move forward into optical. We believe this may foreshadow products from Infinera that would directly competitive with recent DCI and even DC networking solutions from companies like Arista. [Kevin: Looks like if MS, Inphi and others want to “play around” in optical, Infinera can show them how the game is done. Infinera is not about to lose, to anyone. Love this spirit. All-Optical Packet Switching here we come.]

*** Infinite Capacity Engine. Infinera announced a new optical capacity engine that is smaller than a deck of cards and delivers 2.4Tbps over 12k km. Management said that products around this engine are likely to come relatively soon similar to the timing of products around the 500Gbps engine. We believe Infinera is at the stage of designing products rather than finalizing the design of the engine itself.

*** Pay as you go. Dave Welch at Infinera flagged the continuation of the company’s ability to have customers install excess physical capacity in their optical networks and then pay later to turn this capacity on. While the company won’t quantify how much revenue they are receiving from these capacity license activations we believe that the capacity on demand model continues to grow for Infinera and may represent and interesting future path to better margins.

*** Accelerating Engine Cadence. We believe that Infinera is moving its engine design cadence up to every two years with new PIC+DSP solutions coming at a faster clip. We see this as a positive given one deficiency of INFN’s solutions has been sluggishness to deploy new modulations schemes like 16QAM.

Infinera has worked hard to get to where they are. They have superior technology and superior IP. Optical Packet Switching has not been given very much attention. This is where an all optical solution obviates the need for packet level switching vendors. Talk about game changer, this would turn the industry over on its head.

For reference, a quick google search on “All Optical Switch” yields a few scholarly articles: https://www.google.com/webhp?sourceid=chrome-instant&ion…

Below are some of them:

http://www.kt.agh.edu.pl/~iszczesniak/publications/taai.pdf

In optical packet switching the payload (the user data) is switched optically. OPS can be faster, and also cheaper to purchase and maintain than traditional switching with the OEO conversion [12]. OPS hardware could lower power requirements, dissipate less heat and take less space compared to electronic equipment [31].

We can expect OPS to eventually replace traditional electronic switching, because optical network equipment is cheaper to maintain, is more reliable and consume less energy [31] in comparison with their electronic counterparts.

Here’s another one:

http://klamath.stanford.edu/~molinero/thesis/html/pmf_thesis…

This one was for an academic-level talk given by Geoff Bennett in 2001:

http://www.lightreading.com/ethernet-ip/all-optical-switchin…

This is the first of a pair of technology tutorials on all-optical switching by Geoff Bennett, vice president of technology advocacy at Marconi PLC (Nasdaq/London: MONI).

This tutorial covers the all-optical switches themselves – the various types, how they differ from electronic switches, where they sit in networks, what functions they perform, how they’re controlled, and what they can and can’t do.

This is Geoff Bennett today: https://www.linkedin.com/in/geoff-bennett-781254

Also, lest we forget, the IP that came along with Transmode that no one is focusing on. Below are all Transmode whitepapers and articles that got moved onto the Infinera website shortly after the acquisition:

https://www.infinera.com/wp-content/uploads/2015/08/Infinera…

In conventional network architectures, where IP/MPLS operates as a client layer over an optical transport layer, routers are employed throughout the WAN to provide a multitude of packet-centric bandwidth management functions, including packet services termination, aggregation, switching, QoS and transport.

New emerging high-speed packet management functions at 100GbE rates are emerging within next-generation optical transport systems that combine WDM, OTN and packet functionality in a way that provides cable operators with more costeffective tools for optimizing packet traffic, increasing network resource efficiency, and reducing the total cost of the combined network.

This paper compares and contrasts the current PMO with an architectural approach based on packet-aware OTN/optical transport, discusses typical use cases, and presents the benefits this provides operators.

https://www.infinera.com/wp-content/uploads/2015/11/AN_Nativ…

Packet-optical networks have boomed in recent years, driven by the well documented traffic growth in many sectors of the telecoms market and the migration to Ethernet within the networks that support the new services. As a consequence, many network operators have chosen to combine Layer 2 Ethernet functionality with their Layer 1 transport networks to provide a range of new services and to bring economical and operational advantages to these networks.

There are a number of slightly different approaches taken by the vendors in the industry to address this packet-optical evolution, with different technologies being deployed at different places of the network with varying advantages to the network operators. Transmode was an early adopter in the transition, bringing Layer 2 Ethernet functionality directly into metro and regional optical networks. The company’s Native Packet Optical architecture has been deployed by a wide variety of network operators for a broad range of applications.

And this:

http://blog.infinera.com/category/packet-optical/

The list goes on. If you’re interested, here’s a quick google search of “Infinera optical packet switch”:

https://www.google.com/webhp?sourceid=chrome-instant&ion…

Point being - the know how to move packet switching from the ethernet side and put it on the optical side exists in house. They just haven’t put that responsibility onto the PIC because the Arista’s and Cisco’s do that work for them. Well, if Infinera decided to take that part on, then what would be left for the ethernet side? I am no expert, but wouldn’t the ethernet side of the switching function be marginalized to a commodity? I think Infinera could do this if push came to shove. The Inphi solution would be pointless, along with a few other bread and butter technologies and components. We’d be in an all-optical world for transport that converged right into Ethernet. The benefits from that Transmode acquisition may be even better than we suspect. And perhaps those in the know already see this coming as a threat to their longevity.

Thoughts?

Best,
–Kevin

15 Likes

http://hitchhikers.wikia.com/wiki/Infinite_Improbability_Dri…

Sorry. Just had to go there…

5 Likes

INFN would need to get into the internal data center switching business which involves building access layer switches that servers would plug into at speeds like 1G/10G/40G, then aggregation layer switches that bring the access layer switches into a core infrastructure. Those layers are where Cisco sits with it’s bread and butter networking infrastructure.

I am a Datacenter engineer and work in this space so know it fairly well, as far as I know INFN has not entered this space. We had a customer who was running a super computer network that needed 100g internal switching capability, but didn’t like the price point of Cisco’s solutions so I looked at INFN website and didn’t see anything that could accommodate internal data center switching capability, yet.

7 Likes

JC,

What exactly was it that you were looking for that Infn didn’t provide? Also did you look at Anet?

Andy

1 Like

More JPM Notes captured from another source…

This OFC was the most upbeat we have attended. Optical component suppliers we spoke with were universally concerned with whether they will have enough capacity to meet demand. We also believe that Verizon’s 100G deployment is on track for this summer with component orders expected to ramp in May/June. Meanwhile, data center optical demand is also expected to accelerate in Q2 as Tomahawk begins shipping in volume. We see all of this as good news for Ciena, Infinera and Lumentum in our coverage group."

"More on Inphi/Microsoft. After more industry meetings we believe that Microsoft is likely to deploy Inphi’s direct detect solution in shorter links of 10 miles or less associated with large data center campuses they are planning to connect in a mesh. We believe other hyperscalers are tending toward coherent solutions [Kevin: coherent would be the Infinera solution] where they have meshes planned given the relatively low cost relative to the cost of the data centers themselves [Kevin: re. the cost of the DCI equipment is rather insubstantial to begin with in comparison to the cost of the datacenter itself]. We note that we believe this mesh application is new to the DCI world with most current revenue associated with longer haul applications.

So the Inphi solution has gone from 80km on a slide deck to 40km in current “experimental mode” to an anticipated 20km or less in practical use.

Best,
–Kevin

10 Likes

Kevin

Thanks so much for all your posts! Wish I could give you more recs!

Brian

Brian - I’m glad you find them useful. Thank you!

Best,
–Kevin