Switch Where You Can, Route Where You Must

October 16, 2013

Pravin Mahajan - headshotBy Pravin Mahajan

Director, Corporate Marketing

Back in the days of FDDI optical and 10Base5 Thick Ethernet technologies (you remember them, don’t you?), I came across the phrase, “Switch where you can, route where you must.” As a young professional I wondered about it, and applied it to the situation at hand – designing a campus network for a mobile operator. The Synoptics 5000 system was deployed at several places in the network providing Layer 1 and 2 transmission hub and switching functions, while a single Wellfleet Backbone Concentrator Node (BCN) system provided Layer 3 routing (the two companies had actually merged several months before to form Bay Networks). The operator’s campus infrastructure worked perfectly, scaling as traffic volume exploded driven by the mobile revolution.

Route Where 1

I recalled having debates with my manager on the role of OSI Layer 1, 2, 3 technologies in network devices – can or should they be combined, can or should one of them be eliminated, what about processing power, can the store-and-forward technique catch up with cut-through in performance, and more. We in the field close to customers continued deploying validated designs, but R&D labs in companies were feverishly changing the technology landscape and thus attracting Wall Street. Meantime I went back to school and after two years, as I completed my post-graduate degree, the network was still functioning healthily.  New technologies like Fast Ethernet and VLANs were beginning to surface and it was only in early 2000 that the FDDI system was upgraded to Ethernet switching.

I remained in the industry and witnessed tectonic shifts in the telecom technology and business cycles. Now before nostalgia completely overwhelms me, let’s come to the present. The networking world has changed in many ways – we talk terabits, coherent transmission, label switching, and more. But I still reflect on the old adage, “Switch where you can, route where you must,” and wonder on its application today.   What really brought this question back to the surface was the fact that Cisco, a routing technology company, has announced an OTN switching platform that integrates DWDM optics which should be available sometime in 2014.

Here at Infinera we pioneered the integration of optical transmission and digital switching technologies in a single system on the DTN in the mid-2000s. Customers validated this Digital Optical Network architecture and deployed over 2.7 Pb/s of capacity on over 18,000 chassis. They increased efficiency on their network while rapidly provisioning network capacity using standards-based GMPLS control plane and the Infinera Bandwidth Virtualization solution. Last year we extended our portfolio and started shipping the massively scalable DTN-X, which is on a rapid growth trajectory with 39 customers having deployed over three quarters of a Pb/s of super-channel capacity.

The reason for its success is clear – a no performance compromise philosophy while integrating various functions in a single system.  For example, the DTN-X supports 5 Tb/s of capacity across any configuration (100% transmission or 100% switching or any ratio in between) which is scalable to 10 Tb/s per bay and 240 Tb/s in a multi-bay configuration without compromise.

This aspect is quite important as networks are never static – they keep changing due to multiple reasons. One way to understand this is the number of circuits that get deleted or added every year (i.e. churn). We analyzed the data across several global operators for over 18 months, and found some interesting facts – on average, 29% of circuits get churned, and one large content provider had this figure at 79%. We continue to dig deeper into the data for further insights and will share more.


The constant change means that the network must be flexible and simple from an operational standpoint.  Technologies that provision and move circuits efficiently, like switching, need to be widely deployed instead of just using fixed transmission pipes. This is evidenced in the recent global operator survey by Andrew Schmitt of Infonetics on OTN, MPLS and Control Plane Strategies.  A significant number of them (94%) plan to use integrated WDM transmission (Layer 0) and OTN switching (Layer 1) technologies in a single system in the core by 2016. Furthermore, 89% of respondents would like to have OTN and packet switching (i.e. MPLS label switching or Layer 2/2.5) in the same platform by 2016. Andrew further dissects these trends in a brand new whitepaper on multi-layer transport networks. It explains why integrating transmission and switching (Layer 1, Layer 2/2.5) is the right architecture to build an Intelligent Transport Network, with the routing layer staying separate. It also examines why IPoDWDM is not being adopted (hint: layer elimination not integration).

Route Where 2

We have seen many other vendors following in this direction. Ciena and Alcatel-Lucent are working on bringing an integrated switching and DWDM transmission system into their portfolio. And of course, we saw Cisco announce a new system introducing OTN switching for the first time in their portfolio last month.

Today, the provider core architecture is benefiting from the use of an integrated switching solution at most places in the network, with routing being used in a separate layer at only those places needing it.  Some adages stand the test of time and it looks like the one I came across early in my career is continuing to endure. Now if you know who coined it, please share here or drop me a line.

Legal Disclaimer

Leave a Reply

Your email address will not be published. Required fields are marked *