Nice Notes – Take Aways from NGON & DCI Europe 2018
Part II: Just What is Disaggregation?
The disaggregation buzz continues within the optical space, and disaggregation was again a topic of discussion at NGON 2018. However, this year the biggest question about disaggregation was around the definition of disaggregation and the spectrum of possibilities between fully integrated and fully disaggregated. As the disaggregation trend moves from concept to reality, some lessons have been learned and are showing up in equipment and network designs.
Fully integrated systems include the old idea of a “god box” that could handle SONET/SDH, ATM, IP, MPLS, Ethernet, Optical Switching, and a cupholder for good measure. No one seems to be arguing that those complex systems are economical or make sense in a world where the majority of the traffic is now Ethernet-based over an optical layer. However, not everyone agrees that a fully disaggregated system where every piece of hardware has one, and only one, purpose (e.g. single shelf transponders, single shelf ROADM, and single shelf cupholder) is the right way to go in all cases either. Discussions at NGON centered on where the breaking point is for disaggregation, since no one wants to be left behind in the great disaggregation trend.
The AT&T versus Verizon debate on disaggregation always looms large in these discussions, with AT&T saying that disaggregation initiatives like OpenROADM will make them more flexible and Verizon arguing that the complexity of disparate parts can never be less expensive than a combined solution. However, even within those parameters there is room for some gradations of meaning. OpenROADM’s mission isn’t necessarily to break optical components into their smallest parts, its mission is to prove interoperability at the parts level (between transponders, for example). And Verizon’s belief in aggregated systems doesn’t mean that the equipment at the core of their network will have the ability to switch anything other than MPLS packets and wavelengths.
Much of the confusion comes from the requirements of the webscale datacenter operators who legitimately want fixed transponder shelves and open line systems as separate, stackable items. These are the folks that don’t blink about putting in a terabit per second system at one time and have no problem swapping that system out in a few years when something better comes along. For those customers, a pizza box transponder with fixed interfaces makes sense. And at NGON, several industry analysts confirmed that the pizza box transponder is making its way into other networks in specific cases where it is appropriate – generally data center interconnection services, just on a smaller scale. Adding more confusion, the fixed pizza-box concept is now evolving as well, with several vendors touting removable “sleds” that make the hardware more customizable. The term “sled” isn’t really fooling anyone – we all know that those are still cards and slots – but “sled” sounds bigger and less flexible somehow.
Others at NGON spoke of “disaggregated” as the ability to separate optical and packet equipment into separate domains. This concept does not assume that, for example, transponders and amplifiers are necessarily held in separate shelves. It merely assumes that the functions are disaggregated. Therefore, an optical system like ECI’s Apollo could be considered “disaggregated” in this case when no packet cards are inserted. Likewise, a packet system like ECI’s Neptune could be considered “disaggregated” if no optical functions are included. For customers who want to keep their domains separated, this is a legitimate version of disaggregation.
One presentation at NGON used the term “disaggregated” to refer to an Open Line System. Open Line Systems are, to avoid too much complexity, simply optical system that accept alien wavelengths with little or no restrictions. Usually the requirement for open management interfaces is also included in the definition so that a common management platform can be used to monitor and maintain the entire optical network. While Open Line Systems are indeed a valid sub-segment of disaggregation, the entire ecosystem starts to get muddied when we start using a variety of names for the same functionality. It would probably be best to keep OLS and disaggregation separate in the nomenclature, at least in conference presentations. (Feel free to continue to muddy the waters in marketing presentations and sales pitches.)
In the end, discussions generally tended towards an agreement that the term “disaggregation”, like many before it (SDN, POTP, etc.) has now reached a level of acceptance in the industry that is allowing many different flavors of equipment to be called “disaggregated”. The concept is still solid, but operator requirements have meant a move away from strict definitions into more pragmatic solutions. And that’s not a bad thing.